- Changelog for 0.16.4
- Changelog for 0.16.3
- Changelog for 0.16.2
- Changelog for 0.16.1
- Changelog for 0.16.0
- Changelog for 0.15.2
- Changelog for 0.15.1
- Changelog for 0.15.0
- Changelog for 0.14.0
- Changelog for 0.13.0
- Changelog for 0.12.1
- Changelog for 0.12.0
- Changelog for 0.11.0
- Changelog for 0.10.0
- Changelog for 0.9.6
- Changelog for 0.9.5
- Changelog for 0.9.4
- Changelog for 0.9.3
- Changelog for 0.9.2
- Changelog for 0.9.1
- Changelog for 0.9.0
- Changelog for 0.8.3
- Changelog for 0.8.2
- Changelog for 0.8.1
- Changelog for 0.8.0
- Changelog for 0.7.3
- Changelog for 0.7.2
- Changelog for 0.7.1
- Changelog for 0.7.0
- Changelog for 0.6.1
- Changelog for 0.6.0
The following sections list the changes in restic 0.16.4 relevant to restic users. The changes are ordered by importance.
- Fix #4677: Downgrade zstd library to fix rare data corruption at max. compression
- Enh #4529: Add extra verification of data integrity before upload
-
Bugfix #4677: Downgrade zstd library to fix rare data corruption at max. compression
In restic 0.16.3, backups where the compression level was set to
max
(using--compression max
) could in rare and very specific circumstances result in data corruption due to a bug in the library used for compressing data. Restic 0.16.1 and 0.16.2 were not affected.Restic now uses the previous version of the library used to compress data, the same version used by restic 0.16.2. Please note that the
auto
compression level (which restic uses by default) was never affected, and even if you usedmax
compression, chances of being affected by this issue are small.To check a repository for any corruption, run
restic check --read-data
. This will download and verify the whole repository and can be used at any time to completely verify the integrity of a repository. If thecheck
command detects anomalies, follow the suggested steps. -
Enhancement #4529: Add extra verification of data integrity before upload
Hardware issues, or a bug in restic or its dependencies, could previously cause corruption in the files restic created and stored in the repository. Detecting such corruption previously required explicitly running the
check --read-data
orcheck --read-data-subset
commands.To further ensure data integrity, even in the case of hardware issues or software bugs, restic now performs additional verification of the files about to be uploaded to the repository.
These extra checks will increase CPU usage during backups. They can therefore, if absolutely necessary, be disabled using the
--no-extra-verify
global option. Please note that this should be combined with more active checking using the previously mentioned check commands.
The following sections list the changes in restic 0.16.3 relevant to restic users. The changes are ordered by importance.
- Fix #4560: Improve errors for irregular files on Windows
- Fix #4574: Support backup of deduplicated files on Windows again
- Fix #4612: Improve error handling for
rclone
backend - Fix #4624: Correct
restore
progress information if an error occurs - Fix #4626: Improve reliability of restoring large files
-
Bugfix #4560: Improve errors for irregular files on Windows
Since Go 1.21, most filesystem reparse points on Windows are considered to be irregular files. This caused restic to show an
error: invalid node type ""
error message for those files.This error message has now been improved and includes the relevant file path:
error: nodeFromFileInfo path/to/file: unsupported file type "irregular"
. As irregular files are not required to behave like regular files, it is not possible to provide a generic way to back up those files.restic#4560 restic#4620 https://forum.restic.net/t/windows-backup-error-invalid-node-type/6875
-
Bugfix #4574: Support backup of deduplicated files on Windows again
With the official release builds of restic 0.16.1 and 0.16.2, it was not possible to back up files that were deduplicated by the corresponding Windows Server feature. This also applied to restic versions built using Go 1.21.0-1.21.4.
The Go version used to build restic has now been updated to fix this.
-
Bugfix #4612: Improve error handling for
rclone
backendSince restic 0.16.0, if rclone encountered an error while listing files, this could in rare circumstances cause restic to assume that there are no files. Although unlikely, this situation could result in data loss if it were to happen right when the
prune
command is listing existing snapshots.Error handling has now been improved to detect and work around this case.
-
Bugfix #4624: Correct
restore
progress information if an error occursIf an error occurred while restoring a snapshot, this could cause the
restore
progress bar to show incorrect information. In addition, if a data file could not be loaded completely, then errors would also be reported for some already restored files.Error reporting of the
restore
command has now been made more accurate.restic#4624 https://forum.restic.net/t/errors-restoring-with-restic-on-windows-server-s3/6943
-
Bugfix #4626: Improve reliability of restoring large files
In some cases restic failed to restore large files that frequently contain the same file chunk. In combination with certain backends, this could result in network connection timeouts that caused incomplete restores.
Restic now includes special handling for such file chunks to ensure reliable restores.
restic#4626 https://forum.restic.net/t/errors-restoring-with-restic-on-windows-server-s3/6943
The following sections list the changes in restic 0.16.2 relevant to restic users. The changes are ordered by importance.
- Fix #4540: Restore ARMv5 support for ARM binaries
- Fix #4545: Repair documentation build on Read the Docs
-
Bugfix #4540: Restore ARMv5 support for ARM binaries
The official release binaries for restic 0.16.1 were accidentally built to require ARMv7. The build process is now updated to restore support for ARMv5.
Please note that restic 0.17.0 will drop support for ARMv5 and require at least ARMv6.
-
Bugfix #4545: Repair documentation build on Read the Docs
For restic 0.16.1, no documentation was available at https://restic.readthedocs.io/ .
The documentation build process is now updated to work again.
The following sections list the changes in restic 0.16.1 relevant to restic users. The changes are ordered by importance.
- Fix #4513: Make
key list
command honor--no-lock
- Fix #4516: Do not try to load password on command line autocomplete
- Fix #4523: Update zstd library to fix possible data corruption at max. compression
- Chg #4532: Update dependencies and require Go 1.19 or newer
- Enh #229: Show progress bar while loading the index
- Enh #4128: Automatically set
GOMAXPROCS
in resource-constrained containers - Enh #4480: Allow setting REST password and username via environment variables
- Enh #4511: Include inode numbers in JSON output for
find
andls
commands - Enh #4519: Add config option to set SFTP command arguments
-
Bugfix #4513: Make
key list
command honor--no-lock
The
key list
command now supports the--no-lock
options. This allows determining which keys a repo can be accessed by without the need for having write access (e.g., read-only sftp access, filesystem snapshot). -
Bugfix #4516: Do not try to load password on command line autocomplete
The command line autocompletion previously tried to load the repository password. This could cause the autocompletion not to work. Now, this step gets skipped.
-
Bugfix #4523: Update zstd library to fix possible data corruption at max. compression
In restic 0.16.0, backups where the compression level was set to
max
(using--compression max
) could in rare and very specific circumstances result in data corruption due to a bug in the library used for compressing data.Restic now uses the latest version of the library used to compress data, which includes a fix for this issue. Please note that the
auto
compression level (which restic uses by default) was never affected, and even if you usedmax
compression, chances of being affected by this issue were very small.To check a repository for any corruption, run
restic check --read-data
. This will download and verify the whole repository and can be used at any time to completely verify the integrity of a repository. If thecheck
command detects anomalies, follow the suggested steps.To simplify any needed repository repair and minimize data loss, there is also a new and experimental
repair packs
command that salvages all valid data from the affected pack files (seerestic help repair packs
for more information). -
Change #4532: Update dependencies and require Go 1.19 or newer
We have updated all dependencies. Since some libraries require newer Go standard library features, support for Go 1.18 has been dropped, which means that restic now requires at least Go 1.19 to build.
-
Enhancement #229: Show progress bar while loading the index
Restic did not provide any feedback while loading index files. Now, there is a progress bar that shows the index loading progress.
-
Enhancement #4128: Automatically set
GOMAXPROCS
in resource-constrained containersWhen running restic in a Linux container with CPU-usage limits, restic now automatically adjusts
GOMAXPROCS
. This helps to reduce the memory consumption on hosts with many CPU cores. -
Enhancement #4480: Allow setting REST password and username via environment variables
Previously, it was only possible to specify the REST-server username and password in the repository URL, or by using the
--repository-file
option. This meant it was not possible to use authentication in contexts where the repository URL is stored in publicly accessible way.Restic now allows setting the username and password using the
RESTIC_REST_USERNAME
andRESTIC_REST_PASSWORD
variables. -
Enhancement #4511: Include inode numbers in JSON output for
find
andls
commandsRestic used to omit the inode numbers in the JSON messages emitted for nodes by the
ls
command as well as for matches by thefind
command. It now includes those values whenever they are available. -
Enhancement #4519: Add config option to set SFTP command arguments
When using the
sftp
backend, scenarios where a custom identity file was needed for the SSH connection, required the full command to be specified:-o sftp.command='ssh user@host:port -i /ssh/my_private_key -s sftp'
Now, the
-o sftp.args=...
option can be passed to restic to specify custom arguments for the SSH command executed by the SFTP backend. This simplifies the above example to-o sftp.args='-i /ssh/my_private_key'
.
The following sections list the changes in restic 0.16.0 relevant to restic users. The changes are ordered by importance.
- Fix #2565: Support "unlimited" in
forget --keep-*
options - Fix #3311: Support non-UTF8 paths as symlink target
- Fix #4199: Avoid lock refresh issues on slow network connections
- Fix #4274: Improve lock refresh handling after standby
- Fix #4319: Correctly clean up status bar output of the
backup
command - Fix #4333:
generate
andinit
no longer silently ignore unexpected arguments - Fix #4400: Ignore missing folders in
rest
backend - Chg #4176: Fix JSON message type of
scan_finished
for thebackup
command - Chg #4201: Require Go 1.20 for Solaris builds
- Enh #426: Show progress bar during restore
- Enh #719: Add
--retry-lock
option - Enh #1495: Sort snapshots by timestamp in
restic find
- Enh #1759: Add
repair index
andrepair snapshots
commands - Enh #1926: Allow certificate paths to be passed through environment variables
- Enh #2359: Provide multi-platform Docker images
- Enh #2468: Add support for non-global Azure clouds
- Enh #2679: Reduce file fragmentation for local backend
- Enh #3328: Reduce memory usage by up to 25%
- Enh #3397: Improve accuracy of ETA displayed during backup
- Enh #3624: Keep oldest snapshot when there are not enough snapshots
- Enh #3698: Add support for Managed / Workload Identity to
azure
backend - Enh #3871: Support
<snapshot>:<subfolder>
syntax to select subfolders - Enh #3941: Support
--group-by
for backup parent selection - Enh #4130: Cancel current command if cache becomes unusable
- Enh #4159: Add
--human-readable
option tols
andfind
commands - Enh #4188: Include restic version in snapshot metadata
- Enh #4220: Add
jq
binary to Docker image - Enh #4226: Allow specifying region of new buckets in the
gs
backend - Enh #4375: Add support for extended attributes on symlinks
-
Bugfix #2565: Support "unlimited" in
forget --keep-*
optionsRestic would previously forget snapshots that should have been kept when a negative value was passed to the
--keep-*
options. Negative values are now forbidden. To keep all snapshots, the special valueunlimited
is now supported. For example,--keep-monthly unlimited
will keep all monthly snapshots. -
Bugfix #3311: Support non-UTF8 paths as symlink target
Earlier restic versions did not correctly
backup
andrestore
symlinks that contain a non-UTF8 target. Note that this only affected systems that still use a non-Unicode encoding for filesystem paths.The repository format is now extended to add support for such symlinks. Please note that snapshots must have been created with at least restic version 0.16.0 for
restore
to correctly handle non-UTF8 symlink targets when restoring them. -
Bugfix #4199: Avoid lock refresh issues on slow network connections
On network connections with a low upload speed, backups and other operations could fail with the error message
Fatal: failed to refresh lock in time
.This has now been fixed by reworking the lock refresh handling.
-
Bugfix #4274: Improve lock refresh handling after standby
If the restic process was stopped or the host running restic entered standby during a long running operation such as a backup, this previously resulted in the operation failing with
Fatal: failed to refresh lock in time
.This has now been fixed such that restic first checks whether it is safe to continue the current operation and only throws an error if not.
-
Bugfix #4319: Correctly clean up status bar output of the
backup
commandDue to a regression in restic 0.15.2, the status bar of the
backup
command could leave some output behind. This happened if filenames were printed that are wider than the current terminal width. This has now been fixed. -
Bugfix #4333:
generate
andinit
no longer silently ignore unexpected arguments -
Bugfix #4400: Ignore missing folders in
rest
backendIf a repository accessed via the REST backend was missing folders, then restic would fail with an error while trying to list the data in the repository. This has been now fixed.
-
Change #4176: Fix JSON message type of
scan_finished
for thebackup
commandRestic incorrectly set the
message_type
of thescan_finished
message tostatus
instead ofverbose_status
. This has now been corrected so that the messages report the correct type. -
Change #4201: Require Go 1.20 for Solaris builds
Building restic on Solaris now requires Go 1.20, as the library used to access Azure uses the mmap syscall, which is only available on Solaris starting from Go 1.20. All other platforms however continue to build with Go 1.18.
-
Enhancement #426: Show progress bar during restore
The
restore
command now shows a progress report while restoring files.Example:
[0:42] 5.76% 23 files 12.98 MiB, total 3456 files 23.54 GiB
JSON output is now also supported.
restic#426 restic#3413 restic#3627 restic#3991 restic#4314 https://forum.restic.net/t/progress-bar-for-restore/5210
-
Enhancement #719: Add
--retry-lock
optionThis option allows specifying a duration for which restic will wait if the repository is already locked.
-
Enhancement #1495: Sort snapshots by timestamp in
restic find
The
find
command used to print snapshots in an arbitrary order. Restic now prints snapshots sorted by timestamp. -
Enhancement #1759: Add
repair index
andrepair snapshots
commandsThe
rebuild-index
command has been renamed torepair index
. The old name will still work, but is deprecated.When a snapshot was damaged, the only option up to now was to completely forget the snapshot, even if only some unimportant files in it were damaged and other files were still fine.
Restic now has a
repair snapshots
command, which can salvage any non-damaged files and parts of files in the snapshots by removing damaged directories and missing file contents. Please note that the damaged data may still be lost and see the "Troubleshooting" section in the documentation for more details.restic#1759 restic#1714 restic#1798 restic#2334 restic#2876 https://forum.restic.net/t/corrupted-repo-how-to-repair/799 https://forum.restic.net/t/recovery-options-for-damaged-repositories/1571
-
Enhancement #1926: Allow certificate paths to be passed through environment variables
Restic will now read paths to certificates from the environment variables
RESTIC_CACERT
orRESTIC_TLS_CLIENT_CERT
if--cacert
or--tls-client-cert
are not specified. -
Enhancement #2359: Provide multi-platform Docker images
The official Docker images are now built for the architectures linux/386, linux/amd64, linux/arm and linux/arm64.
As an alternative to the Docker Hub, the Docker images are also available on ghcr.io, the GitHub Container Registry.
-
Enhancement #2468: Add support for non-global Azure clouds
The
azure
backend previously only supported storages using the global domaincore.windows.net
. This meant that backups to other domains such as Azure China (core.chinacloudapi.cn
) or Azure Germany (core.cloudapi.de
) were not supported. Restic now allows overriding the global domain using the environment variableAZURE_ENDPOINT_SUFFIX
. -
Enhancement #2679: Reduce file fragmentation for local backend
Before this change, local backend files could become fragmented. Now restic will try to preallocate space for pack files to avoid their fragmentation.
-
Enhancement #3328: Reduce memory usage by up to 25%
The in-memory index has been optimized to be more garbage collection friendly. Restic now defaults to
GOGC=50
to run the Go garbage collector more frequently. -
Enhancement #3397: Improve accuracy of ETA displayed during backup
Restic's
backup
command displayed an ETA that did not adapt when the rate of progress made during the backup changed during the course of the backup.Restic now uses recent progress when computing the ETA. It is important to realize that the estimate may still be wrong, because restic cannot predict the future, but the hope is that the ETA will be more accurate in most cases.
-
Enhancement #3624: Keep oldest snapshot when there are not enough snapshots
The
forget
command now additionally preserves the oldest snapshot if fewer snapshots than allowed by the--keep-*
parameters would otherwise be kept. This maximizes the amount of history kept within the specified limits.restic#3624 restic#4366 https://forum.restic.net/t/keeping-yearly-snapshots-policy-when-backup-began-during-the-year/4670/2
-
Enhancement #3698: Add support for Managed / Workload Identity to
azure
backendRestic now additionally supports authenticating to Azure using Workload Identity or Managed Identity credentials, which are automatically injected in several environments such as a managed Kubernetes cluster.
-
Enhancement #3871: Support
<snapshot>:<subfolder>
syntax to select subfoldersCommands like
diff
orrestore
always worked with the full snapshot. This did not allow comparing only a specific subfolder or only restoring that folder (restore --include subfolder
filters the restored files, but still creates the directories included insubfolder
).The commands
diff
,dump
,ls
andrestore
now support the<snapshot>:<subfolder>
syntax, wheresnapshot
is the ID of a snapshot (or the stringlatest
) andsubfolder
is a path within the snapshot. The commands will then only work with the specified path of the snapshot. Thesubfolder
must be a path to a folder as returned byls
. Two examples:restic restore -t target latest:/some/path
restic diff 12345678:/some/path 90abcef:/some/path
For debugging purposes, the
cat
command now supportscat tree <snapshot>:<subfolder>
to return the directory metadata for the given subfolder. -
Enhancement #3941: Support
--group-by
for backup parent selectionPreviously, the
backup
command by default selected the parent snapshot based on the hostname and the backup targets. When the backup path list changed, thebackup
command was unable to determine a suitable parent snapshot and had to read all files again.The new
--group-by
option for thebackup
command allows filtering snapshots for the parent selection byhost
,paths
andtags
. It defaults tohost,paths
which selects the latest snapshot with hostname and paths matching those of the backup run. This matches the behavior of prior restic versions.The new
--group-by
option should be set to the same value as passed toforget --group-by
. -
Enhancement #4130: Cancel current command if cache becomes unusable
If the cache directory was removed or ran out of space while restic was running, this would previously cause further caching attempts to fail and thereby drastically slow down the command execution. Now, the currently running command is instead canceled.
-
Enhancement #4159: Add
--human-readable
option tols
andfind
commandsPreviously, when using the
-l
option with thels
andfind
commands, the displayed size was always in bytes, without an option for a more human readable format such as MiB or GiB.The new
--human-readable
option will convert longer size values into more human friendly values with an appropriate suffix depending on the output size. For example, a size of14680064
will be shown as14.000 MiB
. -
Enhancement #4188: Include restic version in snapshot metadata
The restic version used to backup a snapshot is now included in its metadata and shown when inspecting a snapshot using
restic cat snapshot <snapshotID>
orrestic snapshots --json
. -
Enhancement #4220: Add
jq
binary to Docker imageThe Docker image now contains
jq
, which can be useful to process JSON data output by restic. -
Enhancement #4226: Allow specifying region of new buckets in the
gs
backendPreviously, buckets used by the Google Cloud Storage backend would always get created in the "us" region. It is now possible to specify the region where a bucket should be created by using the
-o gs.region=us
option. -
Enhancement #4375: Add support for extended attributes on symlinks
Restic now supports extended attributes on symlinks when backing up, restoring, or FUSE-mounting snapshots. This includes, for example, the
security.selinux
xattr on Linux distributions that use SELinux.
The following sections list the changes in restic 0.15.2 relevant to restic users. The changes are ordered by importance.
- Sec #4275: Update golang.org/x/net to address CVE-2022-41723
- Fix #2260: Sanitize filenames printed by
backup
during processing - Fix #4211: Make
dump
interpret--host
and--path
correctly - Fix #4239: Correct number of blocks reported in mount point
- Fix #4253: Minimize risk of spurious filesystem loops with
mount
- Enh #4180: Add release binaries for riscv64 architecture on Linux
- Enh #4219: Upgrade Minio to version 7.0.49
-
Security #4275: Update golang.org/x/net to address CVE-2022-41723
-
Bugfix #2260: Sanitize filenames printed by
backup
during processingThe
backup
command would previously not sanitize the filenames it printed during processing, potentially causing newlines or terminal control characters to mangle the status output or even change the state of a terminal.Filenames are now checked and quoted if they contain non-printable or non-Unicode characters.
-
Bugfix #4211: Make
dump
interpret--host
and--path
correctlyA regression in restic 0.15.0 caused
dump
to confuse its--host=<host>
and--path=<path>
options: it looked for snapshots with paths called<host>
from hosts called<path>
. It now treats the options as intended. -
Bugfix #4239: Correct number of blocks reported in mount point
Restic mount points reported an incorrect number of 512-byte (POSIX standard) blocks for files and links due to a rounding bug. In particular, empty files were reported as taking one block instead of zero.
The rounding is now fixed: the number of blocks reported is the file size (or link target size) divided by 512 and rounded up to a whole number.
-
Bugfix #4253: Minimize risk of spurious filesystem loops with
mount
When a backup contains a directory that has the same name as its parent, say
a/b/b
, and the GNUfind
command was run on this backup in a restic mount,find
would refuse to traverse the lowestb
directory, instead printingFile system loop detected
. This was due to the way the restic mount command generates inode numbers for directories in the mount point.The rule for generating these inode numbers was changed in 0.15.0. It has now been changed again to avoid this issue. A perfect rule does not exist, but the probability of this behavior occurring is now extremely small.
When it does occur, the mount point is not broken, and scripts that traverse the mount point should work as long as they don't rely on inode numbers for detecting filesystem loops.
-
Enhancement #4180: Add release binaries for riscv64 architecture on Linux
Builds for the
riscv64
architecture on Linux are now included in the release binaries. -
Enhancement #4219: Upgrade Minio to version 7.0.49
The upgraded version now allows use of the
ap-southeast-4
region (Melbourne).
The following sections list the changes in restic 0.15.1 relevant to restic users. The changes are ordered by importance.
- Fix #3750: Remove
b2_download_file_by_name: 404
warning from B2 backend - Fix #4147: Make
prune --quiet
not print progress bar - Fix #4163: Make
self-update --output
work with new filename on Windows - Fix #4167: Add missing ETA in
backup
progress bar - Enh #4143: Ignore empty lock files
-
Bugfix #3750: Remove
b2_download_file_by_name: 404
warning from B2 backendIn some cases the B2 backend could print
b2_download_file_by_name: 404: : b2.b2err
warnings. These are only debug messages and can be safely ignored.Restic now uses an updated library for accessing B2, which removes the warning.
-
Bugfix #4147: Make
prune --quiet
not print progress barA regression in restic 0.15.0 caused
prune --quiet
to show a progress bar while deciding how to process each pack files. This has now been fixed. -
Bugfix #4163: Make
self-update --output
work with new filename on WindowsSince restic 0.14.0 the
self-update
command did not work when a custom output filename was specified via the--output
option. This has now been fixed.As a workaround, either use an older restic version to run the self-update or create an empty file with the output filename before updating e.g. using CMD:
type nul > new-file.exe
restic self-update --output new-file.exe
restic#4163 https://forum.restic.net/t/self-update-windows-started-failing-after-release-of-0-15/5836
-
Bugfix #4167: Add missing ETA in
backup
progress barA regression in restic 0.15.0 caused the ETA to be missing from the progress bar displayed by the
backup
command. This has now been fixed. -
Enhancement #4143: Ignore empty lock files
With restic 0.15.0 the checks for stale locks became much stricter than before. In particular, empty or unreadable locks were no longer silently ignored. This made restic to complain with
Load(<lock/1234567812>, 0, 0) returned error, retrying after 552.330144ms: load(<lock/1234567812>): invalid data returned
and fail in the end.The error message is now clarified and the implementation changed to ignore empty lock files which are sometimes created as the result of a failed uploads on some backends.
Please note that unreadable lock files still have to cleaned up manually. To do so, you can run
restic unlock --remove-all
which removes all existing lock files. But first make sure that no other restic process is currently using the repository.
The following sections list the changes in restic 0.15.0 relevant to restic users. The changes are ordered by importance.
- Fix #2015: Make
mount
return exit code 0 after receiving Ctrl-C / SIGINT - Fix #2578: Make
restore
replace existing symlinks - Fix #2591: Don't read password from stdin for
backup --stdin
- Fix #3161: Delete files on Backblaze B2 more reliably
- Fix #3336: Make SFTP backend report no space left on device
- Fix #3567: Improve handling of interrupted syscalls in
mount
command - Fix #3897: Fix stuck
copy
command when-o <backend>.connections=1
- Fix #3918: Correct prune statistics for partially compressed repositories
- Fix #3951: Make
ls
return exit code 1 if snapshot cannot be loaded - Fix #4003: Make
backup
no longer hang on Solaris when seeing a FIFO file - Fix #4016: Support ExFAT-formatted local backends on macOS Ventura
- Fix #4085: Make
init
ignore "Access Denied" errors when creating S3 buckets - Fix #4100: Make
self-update
enabled by default only in release builds - Fix #4103: Don't generate negative UIDs and GIDs in tar files from
dump
- Chg #2724: Include full snapshot ID in JSON output of
backup
- Chg #3929: Make
unlock
display message only when locks were actually removed - Chg #4033: Don't print skipped snapshots by default in
copy
command - Chg #4041: Update dependencies and require Go 1.18 or newer
- Enh #14: Implement
rewrite
command - Enh #79: Restore files with long runs of zeros as sparse files
- Enh #1078: Support restoring symbolic links on Windows
- Enh #1734: Inform about successful retries after errors
- Enh #1866: Improve handling of directories with duplicate entries
- Enh #2134: Support B2 API keys restricted to hiding but not deleting files
- Enh #2152: Make
init
open only one connection for the SFTP backend - Enh #2533: Handle cache corruption on disk and in downloads
- Enh #2715: Stricter repository lock handling
- Enh #2750: Make backup file read concurrency configurable
- Enh #3029: Add support for
credential_process
to S3 backend - Enh #3096: Make
mount
command support macOS using macFUSE 4.x - Enh #3124: Support JSON output for the
init
command - Enh #3899: Optimize prune memory usage
- Enh #3905: Improve speed of parent snapshot detection in
backup
command - Enh #3915: Add compression statistics to the
stats
command - Enh #3925: Provide command completion for PowerShell
- Enh #3931: Allow
backup
file tree scanner to be disabled - Enh #3932: Improve handling of ErrDot errors in rclone and sftp backends
- Enh #3943: Ignore additional/unknown files in repository
- Enh #3955: Improve
backup
performance for small files
-
Bugfix #2015: Make
mount
return exit code 0 after receiving Ctrl-C / SIGINTTo stop the
mount
command, a user has to press Ctrl-C or send a SIGINT signal to restic. This used to cause restic to exit with a non-zero exit code.The exit code has now been changed to zero as the above is the expected way to stop the
mount
command and should therefore be considered successful. -
Bugfix #2578: Make
restore
replace existing symlinksWhen restoring a symlink, restic used to report an error if the target path already existed. This has now been fixed such that the potentially existing target path is first removed before the symlink is restored.
-
Bugfix #2591: Don't read password from stdin for
backup --stdin
The
backup
command when used with--stdin
previously tried to read first the password, then the data to be backed up from standard input. This meant it would often confuse part of the data for the password.From now on, it will instead exit with the message
Fatal: cannot read both password and data from stdin
unless the password is passed in some other way (such as--restic-password-file
,RESTIC_PASSWORD
, etc).To enter the password interactively a password command has to be used. For example on Linux,
mysqldump somedatabase | restic backup --stdin --password-command='sh -c "systemd-ask-password < /dev/tty"'
securely reads the password from the terminal. -
Bugfix #3161: Delete files on Backblaze B2 more reliably
Restic used to only delete the latest version of files stored in B2. In most cases this worked well as there was only a single version of the file. However, due to retries while uploading it is possible for multiple file versions to be stored at B2. This could lead to various problems for files that should have been deleted but still existed.
The implementation has now been changed to delete all versions of files, which doubles the amount of Class B transactions necessary to delete files, but assures that no file versions are left behind.
-
Bugfix #3336: Make SFTP backend report no space left on device
Backing up to an SFTP backend would spew repeated SSH_FX_FAILURE messages when the remote disk was full. Restic now reports "sftp: no space left on device" and exits immediately when it detects this condition.
A fix for this issue was implemented in restic 0.12.1, but unfortunately the fix itself contained a bug that prevented it from taking effect.
-
Bugfix #3567: Improve handling of interrupted syscalls in
mount
commandAccessing restic's FUSE mount could result in "input/output" errors when using programs in which syscalls can be interrupted. This is for example the case for Go programs. This has now been fixed by improved error handling of interrupted syscalls.
-
Bugfix #3897: Fix stuck
copy
command when-o <backend>.connections=1
When running the
copy
command with-o <backend>.connections=1
the command would be infinitely stuck. This has now been fixed. -
Bugfix #3918: Correct prune statistics for partially compressed repositories
In a partially compressed repository, one data blob can exist both in an uncompressed and a compressed version. This caused the
prune
statistics to become inaccurate and e.g. report a too high value for the unused size, such as "unused size after prune: 16777215.991 TiB". This has now been fixed. -
Bugfix #3951: Make
ls
return exit code 1 if snapshot cannot be loadedThe
ls
command used to show a warning and return exit code 0 when failing to load a snapshot. This has now been fixed such that it instead returns exit code 1 (still showing a warning). -
Bugfix #4003: Make
backup
no longer hang on Solaris when seeing a FIFO fileThe
backup
command used to hang on Solaris whenever it encountered a FIFO file (named pipe), due to a bug in the handling of extended attributes. This bug has now been fixed. -
Bugfix #4016: Support ExFAT-formatted local backends on macOS Ventura
ExFAT-formatted disks could not be used as local backends starting from macOS Ventura. Restic commands would fail with an "inappropriate ioctl for device" error. This has now been fixed.
-
Bugfix #4085: Make
init
ignore "Access Denied" errors when creating S3 bucketsIn restic 0.9.0 through 0.13.0, the
init
command ignored some permission errors from S3 backends when trying to check for bucket existence, so that manually created buckets with custom permissions could be used for backups.This feature became broken in 0.14.0, but has now been restored again.
-
Bugfix #4100: Make
self-update
enabled by default only in release buildsThe
self-update
command was previously included by default in all builds of restic as opposed to only in official release builds, even if theselfupdate
tag was not explicitly enabled when building.This has now been corrected, and the
self-update
command is only available if restic was built with-tags selfupdate
(as done for official release builds bybuild.go
). -
Bugfix #4103: Don't generate negative UIDs and GIDs in tar files from
dump
When using a 32-bit build of restic, the
dump
command could in some cases create tar files containing negative UIDs and GIDs, which cannot be read by GNU tar. This corner case especially applies to backups from stdin on Windows.This is now fixed such that
dump
creates valid tar files in these cases too. -
Change #2724: Include full snapshot ID in JSON output of
backup
We have changed the JSON output of the backup command to include the full snapshot ID instead of just a shortened version, as the latter can be ambiguous in some rare cases. To derive the short ID, please truncate the full ID down to eight characters.
-
Change #3929: Make
unlock
display message only when locks were actually removedThe
unlock
command used to print the "successfully removed locks" message whenever it was run, regardless of lock files having being removed or not.This has now been changed such that it only prints the message if any lock files were actually removed. In addition, it also reports the number of removed lock files.
-
Change #4033: Don't print skipped snapshots by default in
copy
commandThe
copy
command used to print each snapshot that was skipped because it already existed in the target repository. The amount of this output could practically bury the list of snapshots that were actually copied.From now on, the skipped snapshots are by default not printed at all, but this can be re-enabled by increasing the verbosity level of the command.
-
Change #4041: Update dependencies and require Go 1.18 or newer
Most dependencies have been updated. Since some libraries require newer language features, support for Go 1.15-1.17 has been dropped, which means that restic now requires at least Go 1.18 to build.
-
Enhancement #14: Implement
rewrite
commandRestic now has a
rewrite
command which allows to rewrite existing snapshots to remove unwanted files. -
Enhancement #79: Restore files with long runs of zeros as sparse files
When using
restore --sparse
, the restorer may now write files containing long runs of zeros as sparse files (also called files with holes), where the zeros are not actually written to disk.How much space is saved by writing sparse files depends on the operating system, file system and the distribution of zeros in the file.
During backup restic still reads the whole file including sparse regions, but with optimized processing speed of sparse regions.
restic#79 restic#3903 restic#2601 restic#3854 https://forum.restic.net/t/sparse-file-support/1264
-
Enhancement #1078: Support restoring symbolic links on Windows
The
restore
command now supports restoring symbolic links on Windows. Because of Windows specific restrictions this is only possible when running restic with theSeCreateSymbolicLinkPrivilege
privilege or as an administrator. -
Enhancement #1734: Inform about successful retries after errors
When a recoverable error is encountered, restic shows a warning message saying that it's retrying, e.g.:
Save(<data/956b9ced99>) returned error, retrying after 357.131936ms: ...
This message can be confusing in that it never clearly states whether the retry is successful or not. This has now been fixed such that restic follows up with a message confirming a successful retry, e.g.:
Save(<data/956b9ced99>) operation successful after 1 retries
-
Enhancement #1866: Improve handling of directories with duplicate entries
If for some reason a directory contains a duplicate entry, the
backup
command would previously fail with anode "path/to/file" already present
ornodes are not ordered got "path/to/file", last "path/to/file"
error.The error handling has been improved to only report a warning in this case. Make sure to check that the filesystem in question is not damaged if you see this!
-
Enhancement #2134: Support B2 API keys restricted to hiding but not deleting files
When the B2 backend does not have the necessary permissions to permanently delete files, it now automatically falls back to hiding files. This allows using restic with an application key which is not allowed to delete files. This can prevent an attacker from deleting backups with such an API key.
To use this feature create an application key without the
deleteFiles
capability. It is recommended to restrict the key to just one bucket. For example using theb2
command line tool:b2 create-key --bucket <bucketName> <keyName> listBuckets,readFiles,writeFiles,listFiles
Alternatively, you can use the S3 backend to access B2, as described in the documentation. In this mode, files are also only hidden instead of being deleted permanently.
-
Enhancement #2152: Make
init
open only one connection for the SFTP backendThe
init
command using the SFTP backend used to connect twice to the repository. This could be inconvenient if the user must enter a password, or causeinit
to fail if the server does not correctly close the first SFTP connection.This has now been fixed by reusing the first/initial SFTP connection opened.
-
Enhancement #2533: Handle cache corruption on disk and in downloads
In rare situations, like for example after a system crash, the data stored in the cache might be corrupted. This could cause restic to fail and required manually deleting the cache.
Restic now automatically removes broken data from the cache, allowing it to recover from such a situation without user intervention. In addition, restic retries downloads which return corrupt data in order to also handle temporary download problems.
-
Enhancement #2715: Stricter repository lock handling
Previously, restic commands kept running even if they failed to refresh their locks in time. This could be a problem e.g. in case the client system running a backup entered the standby power mode while the backup was still in progress (which would prevent the client from refreshing its lock), and after a short delay another host successfully runs
unlock
andprune
on the repository, which would remove all data added by the in-progress backup. If the backup client later continues its backup, even though its lock had expired in the meantime, this would lead to an incomplete snapshot.To address this, lock handling is now much stricter. Commands requiring a lock are canceled if the lock is not refreshed successfully in time. In addition, if a lock file is not readable restic will not allow starting a command. It may be necessary to remove invalid lock files manually or use
unlock --remove-all
. Please make sure that no other restic processes are running concurrently before doing this, however. -
Enhancement #2750: Make backup file read concurrency configurable
The
backup
command now supports a--read-concurrency
option which allows tuning restic for very fast storage like NVMe disks by controlling the number of concurrent file reads during the backup process. -
Enhancement #3029: Add support for
credential_process
to S3 backendRestic now uses a newer library for the S3 backend, which adds support for the
credential_process
option in the AWS credential configuration. -
Enhancement #3096: Make
mount
command support macOS using macFUSE 4.xRestic now uses a different FUSE library for mounting snapshots and making them available as a FUSE filesystem using the
mount
command. This adds support for macFUSE 4.x which can be used to make this work on recent macOS versions. -
Enhancement #3124: Support JSON output for the
init
commandThe
init
command used to ignore the--json
option, but now outputs a JSON message if the repository was created successfully. -
Enhancement #3899: Optimize prune memory usage
The
prune
command needs large amounts of memory in order to determine what to keep and what to remove. This is now optimized to use up to 30% less memory. -
Enhancement #3905: Improve speed of parent snapshot detection in
backup
commandBacking up a large number of files using
--files-from-verbatim
or--files-from-raw
options could require a long time to find the parent snapshot. This has been improved. -
Enhancement #3915: Add compression statistics to the
stats
commandWhen executed with
--mode raw-data
on a repository that supports compression, thestats
command now calculates and displays, for the selected repository or snapshots: the uncompressed size of the data; the compression progress (percentage of data that has been compressed); the compression ratio of the compressed data; the total space saving.It also takes into account both the compressed and uncompressed data if the repository is only partially compressed.
-
Enhancement #3925: Provide command completion for PowerShell
Restic already provided generation of completion files for bash, fish and zsh. Now powershell is supported, too.
-
Enhancement #3931: Allow
backup
file tree scanner to be disabledThe
backup
command walks the file tree in a separate scanner process to find the total size and file/directory count, and uses this to provide an ETA. This can slow down backups, especially of network filesystems.The command now has a new option
--no-scan
which can be used to disable this scanning in order to speed up backups when needed. -
Enhancement #3932: Improve handling of ErrDot errors in rclone and sftp backends
Since Go 1.19, restic can no longer implicitly run relative executables which are found in the current directory (e.g.
rclone
if found in.
). This is a security feature of Go to prevent against running unintended and possibly harmful executables.The error message for this was just "cannot run executable found relative to current directory". This has now been improved to yield a more specific error message, informing the user how to explicitly allow running the executable using the
-o rclone.program
and-o sftp.command
extended options with./
.restic#3932 https://pkg.go.dev/os/exec#hdr-Executables_in_the_current_directory https://go.dev/blog/path-security
-
Enhancement #3943: Ignore additional/unknown files in repository
If a restic repository had additional files in it (not created by restic), commands like
find
andrestore
could become confused and fail with anmultiple IDs with prefix "12345678" found
error. These commands now ignore such additional files.restic#3943 https://forum.restic.net/t/which-protocol-should-i-choose-for-remote-linux-backups/5446/17
-
Enhancement #3955: Improve
backup
performance for small filesWhen backing up small files restic was slower than it could be. In particular this affected backups using maximum compression.
This has been fixed by reworking the internal parallelism of the backup command, making it back up small files around two times faster.
The following sections list the changes in restic 0.14.0 relevant to restic users. The changes are ordered by importance.
- Fix #2248: Support
self-update
on Windows - Fix #3428: List snapshots in backend at most once to resolve snapshot IDs
- Fix #3432: Fix rare 'not found in repository' error for
copy
command - Fix #3681: Fix rclone (shimmed by Scoop) and sftp not working on Windows
- Fix #3685: The
diff
command incorrectly listed some files as added - Fix #3716: Print "wrong password" to stderr instead of stdout
- Fix #3720: Directory sync errors for repositories accessed via SMB
- Fix #3736: The
stats
command miscalculated restore size for multiple snapshots - Fix #3772: Correctly rebuild index for legacy repositories
- Fix #3776: Limit number of key files tested while opening a repository
- Fix #3861: Yield error on invalid policy to
forget
- Chg #1842: Support debug log creation in release builds
- Chg #3295: Deprecate
check --check-unused
and add further checks - Chg #3680: Update dependencies and require Go 1.15 or newer
- Chg #3742: Replace
--repo2
option used byinit
/copy
with--from-repo
- Enh #21: Add compression support
- Enh #1153: Support pruning even when the disk is full
- Enh #2162: Adaptive IO concurrency based on backend connections
- Enh #2291: Allow pack size customization
- Enh #2295: Allow use of SAS token to authenticate to Azure
- Enh #2351: Use config file permissions to control file group access
- Enh #2696: Improve backup speed with many small files
- Enh #2907: Make snapshot directory structure of
mount
command customizable - Enh #2923: Improve speed of
copy
command - Enh #3114: Optimize handling of duplicate blobs in
prune
- Enh #3465: Improve handling of temporary files on Windows
- Enh #3475: Allow limiting IO concurrency for local and SFTP backend
- Enh #3484: Stream data in
check
andprune
commands - Enh #3709: Validate exclude patterns before backing up
- Enh #3729: Display full IDs in
check
warnings - Enh #3773: Optimize memory usage for directories with many files
- Enh #3819: Validate include/exclude patterns before restoring
- Enh #3837: Improve SFTP repository initialization over slow links
-
Bugfix #2248: Support
self-update
on WindowsRestic
self-update
would fail in situations where the operating system locks running binaries, including Windows. The new behavior works around this by renaming the running file and swapping the updated file in place. -
Bugfix #3428: List snapshots in backend at most once to resolve snapshot IDs
Many commands support specifying a list of snapshot IDs which are then used to determine the snapshots to be processed by the command. To resolve snapshot IDs or
latest
, and check that these exist, restic previously listed all snapshots stored in the repository. Depending on the backend this could be a slow and/or expensive operation.Restic now lists the snapshots only once and remembers the result in order to resolve all further snapshot IDs swiftly.
-
Bugfix #3432: Fix rare 'not found in repository' error for
copy
commandIn rare cases
copy
(and other commands) would report thatLoadTree(...)
returned anid [...] not found in repository
error. This could be caused by a backup or copy command running concurrently. The error was only temporary; running the failed restic command a second time as a workaround did resolve the error.This issue has now been fixed by correcting the order in which restic reads data from the repository. It is now guaranteed that restic only loads snapshots for which all necessary data is already available.
-
Bugfix #3681: Fix rclone (shimmed by Scoop) and sftp not working on Windows
In #3602 a fix was introduced to address the problem of
rclone
prematurely exiting when Ctrl+C is pressed on Windows. The solution was to create the subprocess with its console detached from the restic console.However, this solution failed when using
rclone
installed by Scoop or usingsftp
with a passphrase-protected private key. We've now fixed this by using a different approach to prevent Ctrl-C from passing down too early. -
Bugfix #3685: The
diff
command incorrectly listed some files as addedThere was a bug in the
diff
command, causing it to always show files in a removed directory as added. This has now been fixed. -
Bugfix #3716: Print "wrong password" to stderr instead of stdout
If an invalid password was entered, the error message was printed on stdout and not on stderr as intended. This has now been fixed.
-
Bugfix #3720: Directory sync errors for repositories accessed via SMB
On Linux and macOS, accessing a repository via a SMB/CIFS mount resulted in restic failing to save the lock file, yielding the following errors:
Save(<lock/071fe833f0>) returned error, retrying after 552.330144ms: sync /repo/locks: no such file or directory Save(<lock/bf789d7343>) returned error, retrying after 552.330144ms: sync /repo/locks: invalid argument
This has now been fixed by ignoring the relevant error codes.
-
Bugfix #3736: The
stats
command miscalculated restore size for multiple snapshotsSince restic 0.10.0 the restore size calculated by the
stats
command for multiple snapshots was too low. The hardlink detection was accidentally applied across multiple snapshots and thus ignored many files. This has now been fixed. -
Bugfix #3772: Correctly rebuild index for legacy repositories
After running
rebuild-index
on a legacy repository containing mixed pack files (that is, pack files which store both metadata and file data),check
printed warnings likepack 12345678 contained in several indexes: ...
. This warning was not critical, but has now nonetheless been fixed by properly handling mixed pack files while rebuilding the index.Running
prune
for such legacy repositories will also fix the warning by reorganizing the pack files which caused it. -
Bugfix #3776: Limit number of key files tested while opening a repository
Previously, restic tested the password against every key in the repository when opening a repository. The more keys there were in the repository, the slower this operation became.
Restic now tests the password against up to 20 key files in the repository. Alternatively, you can use the
--key-hint=<key ID>
option to specify a specific key file to use instead. -
Bugfix #3861: Yield error on invalid policy to
forget
The
forget
command previously silently ignored invalid/unsupported units in the duration options, such as e.g.--keep-within-daily 2w
.Specifying an invalid/unsupported duration unit now results in an error.
-
Change #1842: Support debug log creation in release builds
Creating a debug log was only possible in debug builds which required users to manually build restic. We changed the release builds to allow creating debug logs by simply setting the environment variable
DEBUG_LOG=logname.log
. -
Change #3295: Deprecate
check --check-unused
and add further checksSince restic 0.12.0, it is expected to still have unused blobs after running
prune
. This made the--check-unused
option of thecheck
command rather useless and tended to confuse users. This option has been deprecated and is now ignored.The
check
command now also warns if a repository is using either the legacy S3 layout or mixed pack files with both tree and data blobs. The latter is known to cause performance problems. -
Change #3680: Update dependencies and require Go 1.15 or newer
We've updated most dependencies. Since some libraries require newer language features we're dropping support for Go 1.14, which means that restic now requires at least Go 1.15 to build.
-
Change #3742: Replace
--repo2
option used byinit
/copy
with--from-repo
The
init
andcopy
commands can read data from another repository. However, confusingly--repo2
referred to the repository from which theinit
command copies parameters, but for thecopy
command--repo2
referred to the copy destination.We've introduced a new option,
--from-repo
, which always refers to the source repository for both commands. The old parameter names have been deprecated but still work. To create a new repository and copy all snapshots to it, the commands are now as follows:restic -r /srv/restic-repo-copy init --from-repo /srv/restic-repo --copy-chunker-params restic -r /srv/restic-repo-copy copy --from-repo /srv/restic-repo
-
Enhancement #21: Add compression support
We've added compression support to the restic repository format. To create a repository using the new format run
init --repository-version 2
. Please note that the repository cannot be read by restic versions prior to 0.14.0.You can configure whether data is compressed with the option
--compression
. It can be set toauto
(the default, which will compress very fast),max
(which will trade backup speed and CPU usage for better compression), oroff
(which disables compression). Each setting is only applied for the current run of restic and does not apply to future runs. The option can also be set via the environment variableRESTIC_COMPRESSION
.To upgrade in place run
migrate upgrade_repo_v2
followed byprune
. See the documentation for more details. The migration checks the repository integrity and upgrades the repository format, but will not change any data. Afterwards, prune will rewrite the metadata to make use of compression.As an alternative you can use the
copy
command to migrate snapshots; First create a new repository usinginit --repository-version 2 --copy-chunker-params --repo2 path/to/old/repo
, and then use thecopy
command to copy all snapshots to the new repository. -
Enhancement #1153: Support pruning even when the disk is full
When running out of disk space it was no longer possible to add or remove data from a repository. To help with recovering from such a deadlock, the prune command now supports an
--unsafe-recover-no-free-space
option to recover from these situations. Make sure to read the documentation first! -
Enhancement #2162: Adaptive IO concurrency based on backend connections
Many commands used hard-coded limits for the number of concurrent operations. This prevented speed improvements by increasing the number of connections used by a backend.
These limits have now been replaced by using the configured number of backend connections instead, which can be controlled using the
-o <backend-name>.connections=5
option. Commands will then automatically scale their parallelism accordingly.To limit the number of CPU cores used by restic, you can set the environment variable
GOMAXPROCS
accordingly. For example to use a single CPU core, useGOMAXPROCS=1
. -
Enhancement #2291: Allow pack size customization
Restic now uses a target pack size of 16 MiB by default. This can be customized using the
--pack-size size
option. Supported pack sizes range between 4 and 128 MiB.It is possible to migrate an existing repository to larger pack files using
prune --repack-small
. This will rewrite every pack file which is significantly smaller than the target size. -
Enhancement #2295: Allow use of SAS token to authenticate to Azure
Previously restic only supported AccountKeys to authenticate to Azure storage accounts, which necessitates giving a significant amount of access.
We added support for Azure SAS tokens which are a more fine-grained and time-limited manner of granting access. Set the
AZURE_ACCOUNT_NAME
andAZURE_ACCOUNT_SAS
environment variables to use a SAS token for authentication. Note that ifAZURE_ACCOUNT_KEY
is set, it will take precedence. -
Enhancement #2351: Use config file permissions to control file group access
Previously files in a local/SFTP repository would always end up with very restrictive access permissions, allowing access only to the owner. This prevented a number of valid use-cases involving groups and ACLs.
We now use the permissions of the config file in the repository to decide whether group access should be given to newly created repository files or not. We arrange for repository files to be created group readable exactly when the repository config file is group readable.
To opt-in to group readable repositories, a simple
chmod -R g+r
or equivalent on the config file can be used. For repositories that should be writable by group members a tad more setup is required, see the docs.Posix ACLs can also be used now that the group permissions being forced to zero no longer masks the effect of ACL entries.
-
Enhancement #2696: Improve backup speed with many small files
We have restructured the backup pipeline to continue reading files while all upload connections are busy. This allows the backup to already prepare the next data file such that the upload can continue as soon as a connection becomes available. This can especially improve the backup performance for high latency backends.
The upload concurrency is now controlled using the
-o <backend-name>.connections=5
option. -
Enhancement #2907: Make snapshot directory structure of
mount
command customizableWe've added the possibility to customize the snapshot directory structure of the
mount
command using templates passed to the--snapshot-template
option. The formatting of snapshots' timestamps is now controlled using--time-template
and supports subdirectories to for example group snapshots by year. Please seerestic help mount
for further details.Characters in tag names which are not allowed in a filename are replaced by underscores
_
. For example a tagfoo/bar
will result in a directory name offoo_bar
. -
Enhancement #2923: Improve speed of
copy
commandThe
copy
command could require a long time to copy snapshots for non-local backends. This has been improved to provide a throughput comparable to therestore
command.Additionally,
copy
now displays a progress bar. -
Enhancement #3114: Optimize handling of duplicate blobs in
prune
Restic
prune
always used to repack all data files containing duplicate blobs. This effectively removed all duplicates during prune. However, as a consequence all these data files were repacked even if the unused repository space threshold could be reached with less work.This is now changed and
prune
works nice and fast even when there are lots of duplicate blobs. -
Enhancement #3465: Improve handling of temporary files on Windows
In some cases restic failed to delete temporary files, causing the current command to fail. This has now been fixed by ensuring that Windows automatically deletes the file. In addition, temporary files are only written to disk when necessary, reducing disk writes.
-
Enhancement #3475: Allow limiting IO concurrency for local and SFTP backend
Restic did not support limiting the IO concurrency / number of connections for accessing repositories stored using the local or SFTP backends. The number of connections is now limited as for other backends, and can be configured via the
-o local.connections=2
and-o sftp.connections=5
options. This ensures that restic does not overwhelm the backend with concurrent IO operations. -
Enhancement #3484: Stream data in
check
andprune
commandsThe commands
check --read-data
andprune
previously downloaded data files into temporary files which could end up being written to disk. This could cause a large amount of data being written to disk.The pack files are now instead streamed, which removes the need for temporary files. Please note that uploads during
backup
andprune
still require temporary files. -
Enhancement #3709: Validate exclude patterns before backing up
Exclude patterns provided via
--exclude
,--iexclude
,--exclude-file
or--iexclude-file
previously weren't validated. As a consequence, invalid patterns resulted in files that were meant to be excluded being backed up.Restic now validates all patterns before running the backup and aborts with a fatal error if an invalid pattern is detected.
-
Enhancement #3729: Display full IDs in
check
warningsWhen running commands to inspect or repair a damaged repository, it is often necessary to supply the full IDs of objects stored in the repository.
The output of
check
now includes full IDs instead of their shortened variant. -
Enhancement #3773: Optimize memory usage for directories with many files
Backing up a directory with hundreds of thousands or more files caused restic to require large amounts of memory. We've now optimized the
backup
command such that it requires up to 30% less memory. -
Enhancement #3819: Validate include/exclude patterns before restoring
Patterns provided to
restore
via--exclude
,--iexclude
,--include
and--iinclude
weren't validated before running the restore. Invalid patterns would result in error messages being printed repeatedly, and possibly unwanted files being restored.Restic now validates all patterns before running the restore, and aborts with a fatal error if an invalid pattern is detected.
-
Enhancement #3837: Improve SFTP repository initialization over slow links
The
init
command, when used on an SFTP backend, now sends multiplemkdir
commands to the backend concurrently. This reduces the waiting times when creating a repository over a very slow connection.
The following sections list the changes in restic 0.13.0 relevant to restic users. The changes are ordered by importance.
- Fix #1106: Never lock repository for
list locks
- Fix #2345: Make cache crash-resistant and usable by multiple concurrent processes
- Fix #2452: Improve error handling of repository locking
- Fix #2738: Don't print progress for
backup --json --quiet
- Fix #3382: Make
check
command honorRESTIC_CACHE_DIR
environment variable - Fix #3488:
rebuild-index
failed if an index file was damaged - Fix #3518: Make
copy
command honor--no-lock
for source repository - Fix #3556: Fix hang with Backblaze B2 on SSL certificate authority error
- Fix #3591: Fix handling of
prune --max-repack-size=0
- Fix #3601: Fix rclone backend prematurely exiting when receiving SIGINT on Windows
- Fix #3619: Avoid choosing parent snapshots newer than time of new snapshot
- Fix #3667: The
mount
command now reports symlinks sizes - Chg #3519: Require Go 1.14 or newer
- Chg #3641: Ignore parent snapshot for
backup --stdin
- Enh #233: Support negative include/exclude patterns
- Enh #1542: Add
--dry-run
/-n
option tobackup
command - Enh #2202: Add upload checksum for Azure, GS, S3 and Swift backends
- Enh #2388: Add warning for S3 if partial credentials are provided
- Enh #2508: Support JSON output and quiet mode for the
diff
command - Enh #2594: Speed up the
restore --verify
command - Enh #2656: Add flag to disable TLS verification for self-signed certificates
- Enh #2816: The
backup
command no longer updates file access times on Linux - Enh #2880: Make
recover
collect only unreferenced trees - Enh #3003: Atomic uploads for the SFTP backend
- Enh #3127: Add xattr (extended attributes) support for Solaris
- Enh #3429: Verify that new or modified keys are stored correctly
- Enh #3436: Improve local backend's resilience to (system) crashes
- Enh #3464: Skip lock creation on
forget
if--no-lock
and--dry-run
- Enh #3490: Support random subset by size in
check --read-data-subset
- Enh #3508: Cache blobs read by the
dump
command - Enh #3511: Support configurable timeout for the rclone backend
- Enh #3541: Improve handling of temporary B2 delete errors
- Enh #3542: Add file mode in symbolic notation to
ls --json
- Enh #3593: Improve
copy
performance by parallelizing IO
-
Bugfix #1106: Never lock repository for
list locks
The
list locks
command previously locked to the repository by default. This had the problem that it wouldn't work for an exclusively locked repository and that the command would also display its own lock file which can be confusing.Now, the
list locks
command never locks the repository. -
Bugfix #2345: Make cache crash-resistant and usable by multiple concurrent processes
The restic cache directory (
RESTIC_CACHE_DIR
) could end up in a broken state in the event of restic (or the OS) crashing. This is now less likely to occur as files are downloaded to a temporary location before being moved to their proper location.This also allows multiple concurrent restic processes to operate on a single repository without conflicts. Previously, concurrent operations could cause segfaults because the processes saw each other's partially downloaded files.
-
Bugfix #2452: Improve error handling of repository locking
Previously, when the lock refresh failed to delete the old lock file, it forgot about the newly created one. Instead it continued trying to delete the old (usually no longer existing) lock file and thus over time lots of lock files accumulated. This has now been fixed.
-
Bugfix #2738: Don't print progress for
backup --json --quiet
Unlike the text output, the
--json
output format still printed progress information even in--quiet
mode. This has now been fixed by always disabling the progress output in quiet mode. -
Bugfix #3382: Make
check
command honorRESTIC_CACHE_DIR
environment variablePreviously, the
check
command didn't honor theRESTIC_CACHE_DIR
environment variable, which caused problems in certain system/usage configurations. This has now been fixed. -
Bugfix #3488:
rebuild-index
failed if an index file was damagedPreviously, the
rebuild-index
command would fail with an error if an index file was damaged or truncated. This has now been fixed.On older restic versions, a (slow) workaround is to use
rebuild-index --read-all-packs
or to manually delete the damaged index. -
Bugfix #3518: Make
copy
command honor--no-lock
for source repositoryThe
copy
command previously did not respect the--no-lock
option for the source repository, causing failures with read-only storage backends. This has now been fixed such that the option is now respected. -
Bugfix #3556: Fix hang with Backblaze B2 on SSL certificate authority error
Previously, if a request failed with an SSL unknown certificate authority error, the B2 backend retried indefinitely and restic would appear to hang.
This has now been fixed and restic instead fails with an error message.
-
Bugfix #3591: Fix handling of
prune --max-repack-size=0
Restic ignored the
--max-repack-size
option when passing a value of 0. This has now been fixed.As a workaround,
--max-repack-size=1
can be used with older versions of restic. -
Bugfix #3601: Fix rclone backend prematurely exiting when receiving SIGINT on Windows
Previously, pressing Ctrl+C in a Windows console where restic was running with rclone as the backend would cause rclone to exit prematurely due to getting a
SIGINT
signal at the same time as restic. Restic would then wait for a long time for time with "unexpected EOF" and "rclone stdio connection already closed" errors.This has now been fixed by restic starting the rclone process detached from the console restic runs in (similar to starting processes in a new process group on Linux), which enables restic to gracefully clean up rclone (which now never gets the
SIGINT
). -
Bugfix #3619: Avoid choosing parent snapshots newer than time of new snapshot
The
backup
command, when a--parent
was not provided, previously chose the most recent matching snapshot as the parent snapshot. However, this didn't make sense when the user passed--time
to create a new snapshot older than the most recent snapshot.Instead,
backup
now chooses the most recent snapshot which is not newer than the snapshot-being-created's timestamp, to avoid any time travel. -
Bugfix #3667: The
mount
command now reports symlinks sizesSymlinks used to have size zero in restic mountpoints, confusing some third-party tools. They now have a size equal to the byte length of their target path, as required by POSIX.
-
Change #3519: Require Go 1.14 or newer
Restic now requires Go 1.14 to build. This allows it to use new standard library features instead of an external dependency.
-
Change #3641: Ignore parent snapshot for
backup --stdin
Restic uses a parent snapshot to speed up directory scanning when performing backups, but this only wasted time and memory when the backup source is stdin (using the
--stdin
option of thebackup
command), since no directory scanning is performed in this case.Snapshots made with
backup --stdin
no longer have a parent snapshot, which allows restic to skip some startup operations and saves a bit of resources.The
--parent
option is still available forbackup --stdin
, but is now ignored. -
Enhancement #233: Support negative include/exclude patterns
If a pattern starts with an exclamation mark and it matches a file that was previously matched by a regular pattern, the match is cancelled. Notably, this can be used with
--exclude-file
to cancel the exclusion of some files.It works similarly to
.gitignore
, with the same limitation; Once a directory is excluded, it is not possible to include files inside the directory.Example of use as an exclude pattern for the
backup
command:$HOME/**/* !$HOME/Documents !$HOME/code !$HOME/.emacs.d !$HOME/games # [...] node_modules *~ *.o *.lo .pyc # [...] $HOME/code/linux/ !$HOME/code/linux/.git
-
Enhancement #1542: Add
--dry-run
/-n
option tobackup
commandTesting exclude filters and other configuration options was error prone as wrong filters could cause files to be uploaded unintentionally. It was also not possible to estimate beforehand how much data would be uploaded.
The
backup
command now has a--dry-run
/-n
option, which performs all the normal steps of a backup without actually writing anything to the repository.Passing -vv will log information about files that would be added, allowing for verification of source and exclusion options before running the real backup.
-
Enhancement #2202: Add upload checksum for Azure, GS, S3 and Swift backends
Previously only the B2 and partially the Swift backends verified the integrity of uploaded (encrypted) files. The verification works by informing the backend about the expected hash of the uploaded file. The backend then verifies the upload and thereby rules out any data corruption during upload.
We have now added upload checksums for the Azure, GS, S3 and Swift backends, which besides integrity checking for uploads also means that restic can now be used to store backups in S3 buckets which have Object Lock enabled.
-
Enhancement #2388: Add warning for S3 if partial credentials are provided
Previously restic did not notify about incomplete credentials when using the S3 backend, instead just reporting access denied.
Restic now checks that both the AWS key ID and secret environment variables are set before connecting to the remote server, and reports an error if not.
-
Enhancement #2508: Support JSON output and quiet mode for the
diff
commandThe
diff
command now supports outputting machine-readable output in JSON format. To enable this, pass the--json
option to the command. To only print the summary and suppress detailed output, pass the--quiet
option. -
Enhancement #2594: Speed up the
restore --verify
commandThe
--verify
option lets therestore
command verify the file content after it has restored a snapshot. The performance of this operation has now been improved by up to a factor of two. -
Enhancement #2656: Add flag to disable TLS verification for self-signed certificates
There is now an
--insecure-tls
global option in restic, which disables TLS verification for self-signed certificates in order to support some development workflows. -
Enhancement #2816: The
backup
command no longer updates file access times on LinuxWhen reading files during backup, restic used to cause the operating system to update the files' access times. Note that this did not apply to filesystems with disabled file access times.
Restic now instructs the operating system not to update the file access time, if the user running restic is the file owner or has root permissions.
-
Enhancement #2880: Make
recover
collect only unreferenced treesPreviously, the
recover
command used to generate a snapshot containing all root trees, even those which were already referenced by a snapshot.This has been improved such that it now only processes trees not already referenced by any snapshot.
-
Enhancement #3003: Atomic uploads for the SFTP backend
The SFTP backend did not upload files atomically. An interrupted upload could leave an incomplete file behind which could prevent restic from accessing the repository. This has now been fixed and uploads in the SFTP backend are done atomically.
-
Enhancement #3127: Add xattr (extended attributes) support for Solaris
Restic now supports xattr for the Solaris operating system.
-
Enhancement #3429: Verify that new or modified keys are stored correctly
When adding a new key or changing the password of a key, restic used to just create the new key (and remove the old one, when changing the password). There was no verification that the new key was stored correctly and works properly. As the repository cannot be decrypted without a valid key file, this could in rare cases cause the repository to become inaccessible.
Restic now checks that new key files actually work before continuing. This can protect against some (rare) cases of hardware or storage problems.
-
Enhancement #3436: Improve local backend's resilience to (system) crashes
Restic now ensures that files stored using the
local
backend are created atomically (that is, files are either stored completely or not at all). This ensures that no incomplete files are left behind even if restic is terminated while writing a file.In addition, restic now tries to ensure that the directory in the repository which contains a newly uploaded file is also written to disk. This can prevent missing files if the system crashes or the disk is not properly unmounted.
-
Enhancement #3464: Skip lock creation on
forget
if--no-lock
and--dry-run
Restic used to silently ignore the
--no-lock
option of theforget
command.It now skips creation of lock file in case both
--dry-run
and--no-lock
are specified. If--no-lock
option is specified without--dry-run
, restic prints a warning message to stderr. -
Enhancement #3490: Support random subset by size in
check --read-data-subset
The
--read-data-subset
option of thecheck
command now supports a third way of specifying the subset to check, namelynS
wheren
is a size in bytes with suffixS
as k/K, m/M, g/G or t/T. -
Enhancement #3508: Cache blobs read by the
dump
commandWhen dumping a file using the
dump
command, restic did not cache blobs in any way, so even consecutive runs of the same blob were loaded from the repository again and again, slowing down the dump.Now, the caching mechanism already used by the
fuse
command is also used by thedump
command. This makes dumping much faster, especially for sparse files. -
Enhancement #3511: Support configurable timeout for the rclone backend
A slow rclone backend could cause restic to time out while waiting for the repository to open. Restic now offers an
-o rclone.timeout
option to make this timeout configurable. -
Enhancement #3541: Improve handling of temporary B2 delete errors
Deleting files on B2 could sometimes fail temporarily, which required restic to retry the delete operation. In some cases the file was deleted nevertheless, causing the retries and ultimately the restic command to fail. This has now been fixed.
-
Enhancement #3542: Add file mode in symbolic notation to
ls --json
The
ls --json
command now provides the file mode in symbolic notation (using thepermissions
key), aligned withfind --json
.restic#3542 restic#3573 https://forum.restic.net/t/restic-ls-understanding-file-mode-with-json/4371
-
Enhancement #3593: Improve
copy
performance by parallelizing IORestic copy previously only used a single thread for copying blobs between repositories, which resulted in limited performance when copying small blobs to/from a high latency backend (i.e. any remote backend, especially b2).
Copying will now use 8 parallel threads to increase the throughput of the copy operation.
The following sections list the changes in restic 0.12.1 relevant to restic users. The changes are ordered by importance.
- Fix #2742: Improve error handling for rclone and REST backend over HTTP2
- Fix #3111: Fix terminal output redirection for PowerShell
- Fix #3184:
backup --quiet
no longer prints status information - Fix #3214: Treat an empty password as a fatal error for repository init
- Fix #3267:
copy
failed to copy snapshots in rare cases - Fix #3296: Fix crash of
check --read-data-subset=x%
run for an empty repository - Fix #3302: Fix
fdopendir: not a directory
error for local backend - Fix #3305: Fix possibly missing backup summary of JSON output in case of error
- Fix #3334: Print
created new cache
message only on a terminal - Fix #3380: Fix crash of
backup --exclude='**'
- Fix #3439: Correctly handle download errors during
restore
- Chg #3247: Empty files now have size of 0 in
ls --json
output - Enh #2780: Add release binaries for s390x architecture on Linux
- Enh #3167: Allow specifying limit of
snapshots
list - Enh #3293: Add
--repository-file2
option toinit
andcopy
command - Enh #3312: Add auto-completion support for fish
- Enh #3336: SFTP backend now checks for disk space
- Enh #3377: Add release binaries for Apple Silicon
- Enh #3414: Add
--keep-within-hourly
option to restic forget - Enh #3426: Optimize read performance of mount command
- Enh #3427:
find --pack
fallback to index if data file is missing - Enh #3456: Support filtering and specifying untagged snapshots
-
Bugfix #2742: Improve error handling for rclone and REST backend over HTTP2
When retrieving data from the rclone / REST backend while also using HTTP2 restic did not detect when no data was returned at all. This could cause for example the
check
command to report the following error:Pack ID does not match, want [...], got e3b0c442
This has been fixed by correctly detecting and retrying the incomplete download.
restic#2742 restic#3453 https://forum.restic.net/t/http2-stream-closed-connection-reset-context-canceled/3743/10
-
Bugfix #3111: Fix terminal output redirection for PowerShell
When redirecting the output of restic using PowerShell on Windows, the output contained terminal escape characters. This has been fixed by properly detecting the terminal type.
In addition, the mintty terminal now shows progress output for the backup command.
-
Bugfix #3184:
backup --quiet
no longer prints status informationA regression in the latest restic version caused the output of
backup --quiet
to contain large amounts of backup progress information when run using an interactive terminal. This is fixed now.A workaround for this bug is to run restic as follows:
restic backup --quiet [..] | cat -
. -
Bugfix #3214: Treat an empty password as a fatal error for repository init
When attempting to initialize a new repository, if an empty password was supplied, the repository would be created but the init command would return an error with a stack trace. Now, if an empty password is provided, it is treated as a fatal error, and no repository is created.
-
Bugfix #3267:
copy
failed to copy snapshots in rare casesThe
copy
command could in rare cases fail with the error messageSaveTree(...) returned unexpected id ...
. This has been fixed.On Linux/BSDs, the error could be caused by backing up symlinks with non-UTF-8 target paths. Note that, due to limitations in the repository format, these are not stored properly and should be avoided if possible.
-
Bugfix #3296: Fix crash of
check --read-data-subset=x%
run for an empty repositoryThe command
restic check --read-data-subset=x%
crashed when run for an empty repository. This has been fixed. -
Bugfix #3302: Fix
fdopendir: not a directory
error for local backendThe
check
,list packs
,prune
andrebuild-index
commands failed for the local backend when thedata
folder in the repository contained files. This has been fixed. -
Bugfix #3305: Fix possibly missing backup summary of JSON output in case of error
When using
--json
output it happened from time to time that the summary output was missing in case an error occurred. This has been fixed. -
Bugfix #3334: Print
created new cache
message only on a terminalThe message
created new cache
was printed even when the output wasn't a terminal. That broke pipingrestic dump
output to tar or zip if cache directory didn't exist. The message is now only printed on a terminal. -
Bugfix #3380: Fix crash of
backup --exclude='**'
The exclude filter
**
, which excludes all files, caused restic to crash. This has been corrected. -
Bugfix #3439: Correctly handle download errors during
restore
Due to a regression in restic 0.12.0, the
restore
command in some cases did not retry download errors and only printed a warning. This has been fixed by retrying incomplete data downloads. -
Change #3247: Empty files now have size of 0 in
ls --json
outputThe
ls --json
command used to omit the sizes of empty files in its output. It now reports a size of zero explicitly for regular files, while omitting the size field for all other types. -
Enhancement #2780: Add release binaries for s390x architecture on Linux
We've added release binaries for Linux using the s390x architecture.
-
Enhancement #3167: Allow specifying limit of
snapshots
listThe
--last
option allowed limiting the output of thesnapshots
command to the latest snapshot for each host. The new--latest n
option allows limiting the output to the latestn
snapshots.This change deprecates the option
--last
in favour of--latest 1
. -
Enhancement #3293: Add
--repository-file2
option toinit
andcopy
commandThe
init
andcopy
command can now be used with the--repository-file2
option or the$RESTIC_REPOSITORY_FILE2
environment variable. These to options are in addition to the--repo2
flag and allow you to read the destination repository from a file.Using both
--repository-file
and--repo2
options resulted in an error for thecopy
orinit
command. The handling of this combination of options has been fixed. A workaround for this issue is to only use--repo
or-r
and--repo2
forinit
orcopy
. -
Enhancement #3312: Add auto-completion support for fish
The
generate
command now supports fish auto completion. -
Enhancement #3336: SFTP backend now checks for disk space
Backing up over SFTP previously spewed multiple generic "failure" messages when the remote disk was full. It now checks for disk space before writing a file and fails immediately with a "no space left on device" message.
-
Enhancement #3377: Add release binaries for Apple Silicon
We've added release binaries for macOS on Apple Silicon (M1).
-
Enhancement #3414: Add
--keep-within-hourly
option to restic forgetThe
forget
command allowed keeping a given number of hourly backups or to keep all backups within a given interval, but it was not possible to specify keeping hourly backups within a given interval.The new
--keep-within-hourly
option now offers this functionality. Similar options for daily/weekly/monthly/yearly are also implemented, the new options are:--keep-within-hourly <1y2m3d4h> --keep-within-daily <1y2m3d4h> --keep-within-weekly <1y2m3d4h> --keep-within-monthly <1y2m3d4h> --keep-within-yearly <1y2m3d4h>
restic#3414 restic#3416 https://forum.restic.net/t/forget-policy/4014/11
-
Enhancement #3426: Optimize read performance of mount command
Reading large files in a mounted repository may be up to five times faster. This improvement primarily applies to repositories stored at a backend that can be accessed with low latency, like e.g. the local backend.
-
Enhancement #3427:
find --pack
fallback to index if data file is missingWhen investigating a repository with missing data files, it might be useful to determine affected snapshots before running
rebuild-index
. Previously,find --pack pack-id
returned no data as it required accessing the data file. Now, if the necessary data is still available in the repository index, it gets retrieved from there.The command now also supports looking up multiple pack files in a single
find
run.restic#3427 https://forum.restic.net/t/missing-packs-not-found/2600
-
Enhancement #3456: Support filtering and specifying untagged snapshots
It was previously not possible to specify an empty tag with the
--tag
and--keep-tag
options. This has now been fixed, such that--tag ''
and--keep-tag ''
now matches snapshots without tags. This allows e.g. thesnapshots
andforget
commands to only operate on untagged snapshots.
The following sections list the changes in restic 0.12.0 relevant to restic users. The changes are ordered by importance.
- Fix #1681: Make
mount
not create missing mount point directory - Fix #1800: Ignore
no data available
filesystem error during backup - Fix #2563: Report the correct owner of directories in FUSE mounts
- Fix #2688: Make
backup
andtag
commands separate tags by comma - Fix #2739: Make the
cat
command respect the--no-lock
option - Fix #3014: Fix sporadic stream reset between rclone and restic
- Fix #3087: The
--use-fs-snapshot
option now works on windows/386 - Fix #3100: Do not require gs bucket permissions when running
init
- Fix #3111: Correctly detect output redirection for
backup
command on Windows - Fix #3151: Don't create invalid snapshots when
backup
is interrupted - Fix #3152: Do not hang until foregrounded when completed in background
- Fix #3166: Improve error handling in the
restore
command - Fix #3232: Correct statistics for overlapping targets
- Fix #3249: Improve error handling in
gs
backend - Chg #3095: Deleting files on Google Drive now moves them to the trash
- Enh #909: Back up mountpoints as empty directories
- Enh #2186: Allow specifying percentage in
check --read-data-subset
- Enh #2433: Make the
dump
command supportzip
format - Enh #2453: Report permanent/fatal backend errors earlier
- Enh #2495: Add option to let
backup
trust mtime without checking ctime - Enh #2528: Add Alibaba/Aliyun OSS support in the
s3
backend - Enh #2706: Configurable progress reports for non-interactive terminals
- Enh #2718: Improve
prune
performance and make it more customizable - Enh #2941: Speed up the repacking step of the
prune
command - Enh #2944: Add
backup
options--files-from-{verbatim,raw}
- Enh #3006: Speed up the
rebuild-index
command - Enh #3048: Add more checks for index and pack files in the
check
command - Enh #3083: Allow usage of deprecated S3
ListObjects
API - Enh #3099: Reduce memory usage of
check
command - Enh #3106: Parallelize scan of snapshot content in
copy
andprune
- Enh #3130: Parallelize reading of locks and snapshots
- Enh #3147: Support additional environment variables for Swift authentication
- Enh #3191: Add release binaries for MIPS architectures
- Enh #3250: Add several more error checks
- Enh #3254: Enable HTTP/2 for backend connections
-
Bugfix #1681: Make
mount
not create missing mount point directoryWhen specifying a non-existent directory as mount point for the
mount
command, restic used to create the specified directory automatically.This has now changed such that restic instead gives an error when the specified directory for the mount point does not exist.
-
Bugfix #1800: Ignore
no data available
filesystem error during backupRestic was unable to backup files on some filesystems, for example certain configurations of CIFS on Linux which return a
no data available
error when reading extended attributes. These errors are now ignored. -
Bugfix #2563: Report the correct owner of directories in FUSE mounts
Restic 0.10.0 changed the FUSE mount to always report the current user as the owner of directories within the FUSE mount, which is incorrect.
This is now changed back to reporting the correct owner of a directory.
-
Bugfix #2688: Make
backup
andtag
commands separate tags by commaRunning
restic backup --tag foo,bar
previously created snapshots with one single tag containing a comma (foo,bar
) instead of two tags (foo
,bar
).Similarly, the
tag
command's--set
,--add
and--remove
options would treatfoo,bar
as one tag instead of two tags. This was inconsistent with other commands and often unexpected when one intendedfoo,bar
to mean two tags.To be consistent in all commands, restic now interprets
foo,bar
to mean two separate tags (foo
andbar
) instead of one tag (foo,bar
) everywhere, including in thebackup
andtag
commands.NOTE: This change might result in unexpected behavior in cases where you use the
forget
command and filter on tags likefoo,bar
. Snapshots previously backed up with--tag foo,bar
will still not match that filter, but snapshots saved from now on will match that filter.To replace
foo,bar
tags withfoo
andbar
tags in old snapshots, you can first generate a list of the relevant snapshots using a command like:Restic snapshots --json --quiet | jq '.[] | select(contains({tags: ["foo,bar"]})) | .id'
And then use
restic tag --set foo --set bar snapshotID [...]
to set the new tags. Please adjust the commands to include real tag names and any additional tags, as well as the list of snapshots to process. -
Bugfix #2739: Make the
cat
command respect the--no-lock
optionThe
cat
command would not respect the--no-lock
flag. This is now fixed. -
Bugfix #3014: Fix sporadic stream reset between rclone and restic
Sometimes when using restic with the
rclone
backend, an error message similar to the following would be printed:Didn't finish writing GET request (wrote 0/xxx): http2: stream closed
It was found that this was caused by restic closing the connection to rclone to soon when downloading data. A workaround has been added which waits for the end of the download before closing the connection.
-
Bugfix #3087: The
--use-fs-snapshot
option now works on windows/386Restic failed to create VSS snapshots on windows/386 with the following error:
GetSnapshotProperties() failed: E_INVALIDARG (0x80070057)
This is now fixed.
-
Bugfix #3100: Do not require gs bucket permissions when running
init
Restic used to require bucket level permissions for the
gs
backend in order to initialize a restic repository.It now allows a
gs
service account to initialize a repository if the bucket does exist and the service account has permissions to write/read to that bucket. -
Bugfix #3111: Correctly detect output redirection for
backup
command on WindowsOn Windows, since restic 0.10.0 the
backup
command did not properly detect when the output was redirected to a file. This caused restic to output terminal control characters. This has been fixed by correcting the terminal detection. -
Bugfix #3151: Don't create invalid snapshots when
backup
is interruptedWhen canceling a backup run at a certain moment it was possible that restic created a snapshot with an invalid "null" tree. This caused
check
and other operations to fail. Thebackup
command now properly handles interruptions and never saves a snapshot when interrupted. -
Bugfix #3152: Do not hang until foregrounded when completed in background
On Linux, when running in the background restic failed to stop the terminal output of the
backup
command after it had completed. This caused restic to hang until moved to the foreground. This has now been fixed.restic#3152 https://forum.restic.net/t/restic-alpine-container-cron-hangs-epoll-pwait/3334
-
Bugfix #3166: Improve error handling in the
restore
commandThe
restore
command used to not print errors while downloading file contents from the repository. It also incorrectly exited with a zero error code even when there were errors during the restore process. This has all been fixed andrestore
now returns with a non-zero exit code when there's an error. -
Bugfix #3232: Correct statistics for overlapping targets
A user reported that restic's statistics and progress information during backup was not correctly calculated when the backup targets (files/dirs to save) overlap. For example, consider a directory
foo
which contains (among others) a filefoo/bar
. Whenrestic backup foo foo/bar
was run, restic counted the size of the filefoo/bar
twice, so the completeness percentage as well as the number of files was wrong. This is now corrected. -
Bugfix #3249: Improve error handling in
gs
backendThe
gs
backend did not notice when the last step of completing a file upload failed. Under rare circumstances, this could cause missing files in the backup repository. This has now been fixed. -
Change #3095: Deleting files on Google Drive now moves them to the trash
When deleting files on Google Drive via the
rclone
backend, restic used to bypass the trash folder required that one used the-o rclone.args
option to enable usage of the trash folder. This ensured that deleted files in Google Drive were not kept indefinitely in the trash folder. However, since Google Drive's trash retention policy changed to deleting trashed files after 30 days, this is no longer needed.Restic now leaves it up to rclone and its configuration to use or not use the trash folder when deleting files. The default is to use the trash folder, as of rclone 1.53.2. To re-enable the restic 0.11 behavior, set the
RCLONE_DRIVE_USE_TRASH
environment variable or change the rclone configuration. See the rclone documentation for more details. -
Enhancement #909: Back up mountpoints as empty directories
When the
--one-file-system
option is specified torestic backup
, it ignores all file systems mounted below one of the target directories. This means that when a snapshot is restored, users needed to manually recreate the mountpoint directories.Restic now backs up mountpoints as empty directories and therefore implements the same approach as
tar
. -
Enhancement #2186: Allow specifying percentage in
check --read-data-subset
We've enhanced the
check
command's--read-data-subset
option to also accept a percentage (e.g.2.5%
or10%
). This will check the given percentage of pack files (which are randomly selected on each run). -
Enhancement #2433: Make the
dump
command supportzip
formatPreviously, restic could dump the contents of a whole folder structure only in the
tar
format. Thedump
command now has a new flag to change output format tozip
. Just pass--archive zip
as an option torestic dump
. -
Enhancement #2453: Report permanent/fatal backend errors earlier
When encountering errors in reading from or writing to storage backends, restic retries the failing operation up to nine times (for a total of ten attempts). It used to retry all backend operations, but now detects some permanent error conditions so that it can report fatal errors earlier.
Permanent failures include local disks being full, SSH connections dropping and permission errors.
-
Enhancement #2495: Add option to let
backup
trust mtime without checking ctimeThe
backup
command used to require that bothctime
andmtime
of a file matched with a previously backed up version to determine that the file was unchanged. In other words, if eitherctime
ormtime
of the file had changed, it would be considered changed and restic would read the file's content again to back up the relevant (changed) parts of it.The new option
--ignore-ctime
makes restic look atmtime
only, such thatctime
changes for a file does not cause restic to read the file's contents again.The check for both
ctime
andmtime
was introduced in restic 0.9.6 to make backups more reliable in the face of programs that resetmtime
(some Unix archivers do that), but it turned out to often be expensive because it made restic read file contents even if only the metadata (owner, permissions) of a file had changed. The new--ignore-ctime
option lets the user restore the 0.9.5 behavior when needed. The existing--ignore-inode
option already turned off this behavior, but also removed a different check.Please note that changes in files' metadata are still recorded, regardless of the command line options provided to the backup command.
-
Enhancement #2528: Add Alibaba/Aliyun OSS support in the
s3
backendA new extended option
s3.bucket-lookup
has been added to support Alibaba/Aliyun OSS in thes3
backend. The option can be set to one of the following values:auto
- Existing behaviour -dns
- Use DNS style bucket access -path
- Use path style bucket access
To make the
s3
backend work with Alibaba/Aliyun OSS you must sets3.bucket-lookup
todns
and set thes3.region
parameter. For example:Restic -o s3.bucket-lookup=dns -o s3.region=oss-eu-west-1 -r s3:https://oss-eu-west-1.aliyuncs.com/bucketname init
Note that
s3.region
must be set, otherwise the MinIO SDK tries to look it up and it seems that Alibaba doesn't support that properly. -
Enhancement #2706: Configurable progress reports for non-interactive terminals
The
backup
,check
andprune
commands never printed any progress reports on non-interactive terminals. This behavior is now configurable using theRESTIC_PROGRESS_FPS
environment variable. Use for example a value of1
for an update every second, or0.01666
for an update every minute.The
backup
command now also prints the current progress when restic receives aSIGUSR1
signal.Setting the
RESTIC_PROGRESS_FPS
environment variable or sending aSIGUSR1
signal prints a status report even when--quiet
was specified. -
Enhancement #2718: Improve
prune
performance and make it more customizableThe
prune
command is now much faster. This is especially the case for remote repositories or repositories with not much data to remove. Also the memory usage of theprune
command is now reduced.Restic used to rebuild the index from scratch after pruning. This could lead to missing packs in the index in some cases for eventually consistent backends such as e.g. AWS S3. This behavior is now changed and the index rebuilding uses the information already known by
prune
.By default, the
prune
command no longer removes all unused data. This behavior can be fine-tuned by new options, like the acceptable amount of unused space or the maximum size of data to reorganize. For more details, please see https://restic.readthedocs.io/en/stable/060_forget.html .Moreover,
prune
now accepts the--dry-run
option and also runningforget --dry-run --prune
will show whatprune
would do.This enhancement also fixes several open issues, e.g.: - restic#1140 - restic#1599 - restic#1985 - restic#2112 - restic#2227 - restic#2305
-
Enhancement #2941: Speed up the repacking step of the
prune
commandThe repack step of the
prune
command, which moves still used file parts into new pack files such that the old ones can be garbage collected later on, now processes multiple pack files in parallel. This is especially beneficial for high latency backends or when using a fast network connection. -
Enhancement #2944: Add
backup
options--files-from-{verbatim,raw}
The new
backup
options--files-from-verbatim
and--files-from-raw
read a list of files to back up from a file. Unlike the existing--files-from
option, these options do not interpret the listed filenames as glob patterns; instead, whitespace in filenames is preserved as-is and no pattern expansion is done. Please see the documentation for specifics.These new options are highly recommended over
--files-from
, when using a script to generate the list of files to back up. -
Enhancement #3006: Speed up the
rebuild-index
commandWe've optimized the
rebuild-index
command. Now, existing index entries are used to minimize the number of pack files that must be read. This speeds up the index rebuild a lot.Additionally, the option
--read-all-packs
has been added, implementing the previous behavior. -
Enhancement #3048: Add more checks for index and pack files in the
check
commandThe
check
command run with the--read-data
or--read-data-subset
options used to only verify only the pack file content - it did not check if the blobs within the pack are correctly contained in the index.A check for the latter is now in place, which can print the following error:
Blob ID is not contained in index or position is incorrect
Another test is also added, which compares pack file sizes computed from the index and the pack header with the actual file size. This test is able to detect truncated pack files.
If the index is not correct, it can be rebuilt by using the
rebuild-index
command.Having added these tests,
restic check
is now able to detect non-existing blobs which are wrongly referenced in the index. This situation could have lead to missing data. -
Enhancement #3083: Allow usage of deprecated S3
ListObjects
APISome S3 API implementations, e.g. Ceph before version 14.2.5, have a broken
ListObjectsV2
implementation which causes problems for restic when using their API endpoints. When a broken server implementation is used, restic prints errors similar to the following:List() returned error: Truncated response should have continuation token set
As a temporary workaround, restic now allows using the older
ListObjects
endpoint by setting thes3.list-objects-v1
extended option, for instance:Restic -o s3.list-objects-v1=true snapshots
Please note that this option may be removed in future versions of restic.
-
Enhancement #3099: Reduce memory usage of
check
commandThe
check
command now requires less memory if it is run without the--check-unused
option. -
Enhancement #3106: Parallelize scan of snapshot content in
copy
andprune
The
copy
andprune
commands used to traverse the directories of snapshots one by one to find used data. This snapshot traversal is now parallized which can speed up this step several times.In addition the
check
command now reports how many snapshots have already been processed. -
Enhancement #3130: Parallelize reading of locks and snapshots
Restic used to read snapshots sequentially. For repositories containing many snapshots this slowed down commands which have to read all snapshots.
Now the reading of snapshots is parallelized. This speeds up for example
prune
,backup
and other commands that search for snapshots with certain properties or which have to find thelatest
snapshot.The speed up also applies to locks stored in the backup repository.
-
Enhancement #3147: Support additional environment variables for Swift authentication
The
swift
backend now supports the following additional environment variables for passing authentication details to restic:OS_USER_ID
,OS_USER_DOMAIN_ID
,OS_PROJECT_DOMAIN_ID
andOS_TRUST_ID
Depending on the
openrc
configuration file these might be required when the user and project domains differ from one another. -
Enhancement #3191: Add release binaries for MIPS architectures
We've added a few new architectures for Linux to the release binaries:
mips
,mipsle
,mips64
, andmip64le
. MIPS is mostly used for low-end embedded systems. -
Enhancement #3250: Add several more error checks
We've added a lot more error checks in places where errors were previously ignored (as hinted by the static analysis program
errcheck
viagolangci-lint
). -
Enhancement #3254: Enable HTTP/2 for backend connections
Go's HTTP library usually automatically chooses between HTTP/1.x and HTTP/2 depending on what the server supports. But for compatibility this mechanism is disabled if DialContext is used (which is the case for restic). This change allows restic's HTTP client to negotiate HTTP/2 if supported by the server.
The following sections list the changes in restic 0.11.0 relevant to restic users. The changes are ordered by importance.
- Fix #1212: Restore timestamps and permissions on intermediate directories
- Fix #1756: Mark repository files as read-only when using the local backend
- Fix #2241: Hide password in REST backend repository URLs
- Fix #2319: Correctly dump directories into tar files
- Fix #2491: Don't require
self-update --output
placeholder file - Fix #2834: Fix rare cases of backup command hanging forever
- Fix #2938: Fix manpage formatting
- Fix #2942: Make --exclude-larger-than handle disappearing files
- Fix #2951: Restic generate, help and self-update no longer check passwords
- Fix #2979: Make snapshots --json output [] instead of null when no snapshots
- Enh #340: Add support for Volume Shadow Copy Service (VSS) on Windows
- Enh #1458: New option --repository-file
- Enh #2849: Authenticate to Google Cloud Storage with access token
- Enh #2969: Optimize check for unchanged files during backup
- Enh #2978: Warn if parent snapshot cannot be loaded during backup
-
Bugfix #1212: Restore timestamps and permissions on intermediate directories
When using the
--include
option of the restore command, restic restored timestamps and permissions only on directories selected by the include pattern. Intermediate directories, which are necessary to restore files located in sub- directories, were created with default permissions. We've fixed the restore command to restore timestamps and permissions for these directories as well. -
Bugfix #1756: Mark repository files as read-only when using the local backend
Files stored in a local repository were marked as writeable on the filesystem for non-Windows systems, which did not prevent accidental file modifications outside of restic. In addition, the local backend did not work with certain filesystems and network mounts which do not permit modifications of file permissions.
Restic now marks files stored in a local repository as read-only on the filesystem on non-Windows systems. The error handling is improved to support more filesystems.
-
Bugfix #2241: Hide password in REST backend repository URLs
When using a password in the REST backend repository URL, the password could in some cases be included in the output from restic, e.g. when initializing a repo or during an error.
The password is now replaced with "***" where applicable.
-
Bugfix #2319: Correctly dump directories into tar files
The dump command previously wrote directories in a tar file in a way which can cause compatibility problems. This caused, for example, 7zip on Windows to not open tar files containing directories. In addition it was not possible to dump directories with extended attributes. These compatibility problems are now corrected.
In addition, a tar file now includes the name of the owner and group of a file.
-
Bugfix #2491: Don't require
self-update --output
placeholder filerestic self-update --output /path/to/new-restic
used to require that new-restic was an existing file, to be overwritten. Now it's possible to download an updated restic binary to a new path, without first having to create a placeholder file. -
Bugfix #2834: Fix rare cases of backup command hanging forever
We've fixed an issue with the backup progress reporting which could cause restic to hang forever right before finishing a backup.
-
Bugfix #2938: Fix manpage formatting
The manpage formatting in restic v0.10.0 was garbled, which is fixed now.
-
Bugfix #2942: Make --exclude-larger-than handle disappearing files
There was a small bug in the backup command's --exclude-larger-than option where files that disappeared between scanning and actually backing them up to the repository caused a panic. This is now fixed.
-
Bugfix #2951: Restic generate, help and self-update no longer check passwords
The commands
restic cache
,generate
,help
andself-update
don't need passwords, but they previously did run the RESTIC_PASSWORD_COMMAND (if set in the environment), prompting users to authenticate for no reason. They now skip running the password command. -
Bugfix #2979: Make snapshots --json output [] instead of null when no snapshots
Restic previously output
null
instead of[]
for the--json snapshots
command, when there were no snapshots in the repository. This caused some minor problems when parsing the output, but is now fixed such that[]
is output when the list of snapshots is empty. -
Enhancement #340: Add support for Volume Shadow Copy Service (VSS) on Windows
Volume Shadow Copy Service allows read access to files that are locked by another process using an exclusive lock through a filesystem snapshot. Restic was unable to backup those files before. This update enables backing up these files.
This needs to be enabled explicitely using the --use-fs-snapshot option of the backup command.
-
Enhancement #1458: New option --repository-file
We've added a new command-line option --repository-file as an alternative to -r. This allows to read the repository URL from a file in order to prevent certain types of information leaks, especially for URLs containing credentials.
-
Enhancement #2849: Authenticate to Google Cloud Storage with access token
When using the GCS backend, it is now possible to authenticate with OAuth2 access tokens instead of a credentials file by setting the GOOGLE_ACCESS_TOKEN environment variable.
-
Enhancement #2969: Optimize check for unchanged files during backup
During a backup restic skips processing files which have not changed since the last backup run. Previously this required opening each file once which can be slow on network filesystems. The backup command now checks for file changes before opening a file. This considerably reduces the time to create a backup on network filesystems.
-
Enhancement #2978: Warn if parent snapshot cannot be loaded during backup
During a backup restic uses the parent snapshot to check whether a file was changed and has to be backed up again. For this check the backup has to read the directories contained in the old snapshot. If a tree blob cannot be loaded, restic now warns about this problem with the backup repository.
The following sections list the changes in restic 0.10.0 relevant to restic users. The changes are ordered by importance.
- Fix #1863: Report correct number of directories processed by backup
- Fix #2254: Fix tar issues when dumping
/
- Fix #2281: Handle format verbs like '%' properly in
find
output - Fix #2298: Do not hang when run as a background job
- Fix #2389: Fix mangled json output of backup command
- Fix #2390: Refresh lock timestamp
- Fix #2429: Backup --json reports total_bytes_processed as 0
- Fix #2469: Fix incorrect bytes stats in
diff
command - Fix #2518: Do not crash with Synology NAS sftp server
- Fix #2531: Fix incorrect size calculation in
stats --mode restore-size
- Fix #2537: Fix incorrect file counts in
stats --mode restore-size
- Fix #2592: SFTP backend supports IPv6 addresses
- Fix #2607: Honor RESTIC_CACHE_DIR environment variable on Mac and Windows
- Fix #2668: Don't abort the stats command when data blobs are missing
- Fix #2674: Add stricter prune error checks
- Fix #2899: Fix possible crash in the progress bar of check --read-data
- Chg #1597: Honor the --no-lock flag in the mount command
- Chg #2482: Remove vendored dependencies
- Chg #2546: Return exit code 3 when failing to backup all source data
- Chg #2600: Update dependencies, require Go >= 1.13
- Enh #323: Add command for copying snapshots between repositories
- Enh #551: Use optimized library for hash calculation of file chunks
- Enh #1570: Support specifying multiple host flags for various commands
- Enh #1680: Optimize
restic mount
- Enh #2072: Display snapshot date when using
restic find
- Enh #2175: Allow specifying user and host when creating keys
- Enh #2195: Simplify and improve restore performance
- Enh #2277: Add support for ppc64le
- Enh #2328: Improve speed of check command
- Enh #2395: Ignore sync errors when operation not supported by local filesystem
- Enh #2423: Support user@domain parsing as user
- Enh #2427: Add flag
--iexclude-file
to backup command - Enh #2569: Support excluding files by their size
- Enh #2571: Self-heal missing file parts during backup of unchanged files
- Enh #2576: Improve the chunking algorithm
- Enh #2598: Improve speed of diff command
- Enh #2599: Slightly reduce memory usage of prune and stats commands
- Enh #2733: S3 backend: Add support for WebIdentityTokenFile
- Enh #2773: Optimize handling of new index entries
- Enh #2781: Reduce memory consumption of in-memory index
- Enh #2786: Optimize
list blobs
command - Enh #2790: Optimized file access in restic mount
- Enh #2840: Speed-up file deletion in forget, prune and rebuild-index
- Enh #2858: Support filtering snapshots by tag and path in the stats command
-
Bugfix #1863: Report correct number of directories processed by backup
The directory statistics calculation was fixed to report the actual number of processed directories instead of always zero.
-
Bugfix #2254: Fix tar issues when dumping
/
We've fixed an issue with dumping either
/
or files on the first sublevel e.g./foo
to tar. This also fixes tar dumping issues on Windows where this issue could also happen. -
Bugfix #2281: Handle format verbs like '%' properly in
find
outputThe JSON or "normal" output of the
find
command can now deal with file names that contain substrings which the Golangfmt
package considers "format verbs" like%s
. -
Bugfix #2298: Do not hang when run as a background job
Restic did hang on exit while restoring the terminal configuration when it was started as a background job, for example using
restic ... &
. This has been fixed by only restoring the terminal configuration when restic is interrupted while reading a password from the terminal. -
Bugfix #2389: Fix mangled json output of backup command
We've fixed a race condition in the json output of the backup command that could cause multiple lines to get mixed up. We've also ensured that the backup summary is printed last.
-
Bugfix #2390: Refresh lock timestamp
Long-running operations did not refresh lock timestamp, resulting in locks becoming stale. This is now fixed.
-
Bugfix #2429: Backup --json reports total_bytes_processed as 0
We've fixed the json output of total_bytes_processed. The non-json output was already fixed with pull request #2138 but left the json output untouched.
-
Bugfix #2469: Fix incorrect bytes stats in
diff
commandIn some cases, the wrong number of bytes (e.g. 16777215.998 TiB) were reported by the
diff
command. This is now fixed. -
Bugfix #2518: Do not crash with Synology NAS sftp server
It was found that when restic is used to store data on an sftp server on a Synology NAS with a relative path (one which does not start with a slash), it may go into an endless loop trying to create directories on the server. We've fixed this bug by using a function in the sftp library instead of our own implementation.
The bug was discovered because the Synology sftp server behaves erratic with non-absolute path (e.g.
home/restic-repo
). This can be resolved by just using an absolute path instead (/home/restic-repo
). We've also added a paragraph in the FAQ. -
Bugfix #2531: Fix incorrect size calculation in
stats --mode restore-size
The restore-size mode of stats was counting hard-linked files as if they were independent.
-
Bugfix #2537: Fix incorrect file counts in
stats --mode restore-size
The restore-size mode of stats was failing to count empty directories and some files with hard links.
-
Bugfix #2592: SFTP backend supports IPv6 addresses
The SFTP backend now supports IPv6 addresses natively, without relying on aliases in the external SSH configuration.
-
Bugfix #2607: Honor RESTIC_CACHE_DIR environment variable on Mac and Windows
On Mac and Windows, the RESTIC_CACHE_DIR environment variable was ignored. This variable can now be used on all platforms to set the directory where restic stores caches.
-
Bugfix #2668: Don't abort the stats command when data blobs are missing
Runing the stats command in the blobs-per-file mode on a repository with missing data blobs previously resulted in a crash.
-
Bugfix #2674: Add stricter prune error checks
Additional checks were added to the prune command in order to improve resiliency to backend, hardware and/or networking issues. The checks now detect a few more cases where such outside factors could potentially cause data loss.
-
Bugfix #2899: Fix possible crash in the progress bar of check --read-data
We've fixed a possible crash while displaying the progress bar for the check --read-data command. The crash occurred when the length of the progress bar status exceeded the terminal width, which only happened for very narrow terminal windows.
restic#2899 https://forum.restic.net/t/restic-rclone-pcloud-connection-issues/2963/15
-
Change #1597: Honor the --no-lock flag in the mount command
The mount command now does not lock the repository if given the --no-lock flag. This allows to mount repositories which are archived on a read only backend/filesystem.
-
Change #2482: Remove vendored dependencies
We've removed the vendored dependencies (in the subdir
vendor/
). When building restic, the Go compiler automatically fetches the dependencies. It will also cryptographically verify that the correct code has been fetched by using the hashes ingo.sum
(see the link to the documentation below).restic#2482 https://golang.org/cmd/go/#hdr-Module_downloading_and_verification
-
Change #2546: Return exit code 3 when failing to backup all source data
The backup command used to return a zero exit code as long as a snapshot could be created successfully, even if some of the source files could not be read (in which case the snapshot would contain the rest of the files).
This made it hard for automation/scripts to detect failures/incomplete backups by looking at the exit code. Restic now returns the following exit codes for the backup command:
- 0 when the command was successful - 1 when there was a fatal error (no snapshot created) - 3 when some source data could not be read (incomplete snapshot created)
-
Change #2600: Update dependencies, require Go >= 1.13
Restic now requires Go to be at least 1.13. This allows simplifications in the build process and removing workarounds.
This is also probably the last version of restic still supporting mounting repositories via fuse on macOS. The library we're using for fuse does not support macOS any more and osxfuse is not open source any more.
bazil/fuse#224 osxfuse/osxfuse#590 restic#2600 restic#2852 restic#2927
-
Enhancement #323: Add command for copying snapshots between repositories
We've added a copy command, allowing you to copy snapshots from one repository to another.
Note that this process will have to read (download) and write (upload) the entire snapshot(s) due to the different encryption keys used on the source and destination repository. Also, the transferred files are not re-chunked, which may break deduplication between files already stored in the destination repo and files copied there using this command.
To fully support deduplication between repositories when the copy command is used, the init command now supports the
--copy-chunker-params
option, which initializes the new repository with identical parameters for splitting files into chunks as an already existing repository. This allows copied snapshots to be equally deduplicated in both repositories. -
Enhancement #551: Use optimized library for hash calculation of file chunks
We've switched the library used to calculate the hashes of file chunks, which are used for deduplication, to the optimized Minio SHA-256 implementation.
Depending on the CPU it improves the hashing throughput by 10-30%. Modern x86 CPUs with the SHA Extension should be about two to three times faster.
-
Enhancement #1570: Support specifying multiple host flags for various commands
Previously commands didn't take more than one
--host
or-H
argument into account, which could be limiting with e.g. theforget
command.The
dump
,find
,forget
,ls
,mount
,restore
,snapshots
,stats
andtag
commands will now take into account multiple--host
and-H
flags. -
Enhancement #1680: Optimize
restic mount
We've optimized the FUSE implementation used within restic.
restic mount
is now more responsive and uses less memory. -
Enhancement #2072: Display snapshot date when using
restic find
Added the respective snapshot date to the output of
restic find
. -
Enhancement #2175: Allow specifying user and host when creating keys
When adding a new key to the repository, the username and hostname for the new key can be specified on the command line. This allows overriding the defaults, for example if you would prefer to use the FQDN to identify the host or if you want to add keys for several different hosts without having to run the key add command on those hosts.
-
Enhancement #2195: Simplify and improve restore performance
Significantly improves restore performance of large files (i.e. 50M+): restic#2074 https://forum.restic.net/t/restore-using-rclone-gdrive-backend-is-slow/1112/8 https://forum.restic.net/t/degraded-restore-performance-s3-backend/1400
Fixes "not enough cache capacity" error during restore: restic#2244
NOTE: This new implementation does not guarantee order in which blobs are written to the target files and, for example, the last blob of a file can be written to the file before any of the preceding file blobs. It is therefore possible to have gaps in the data written to the target files if restore fails or interrupted by the user.
The implementation will try to preallocate space for the restored files on the filesystem to prevent file fragmentation. This ensures good read performance for large files, like for example VM images. If preallocating space is not supported by the filesystem, then this step is silently skipped.
-
Enhancement #2277: Add support for ppc64le
Adds support for ppc64le, the processor architecture from IBM.
-
Enhancement #2328: Improve speed of check command
We've improved the check command to traverse trees only once independent of whether they are contained in multiple snapshots. The check command is now much faster for repositories with a large number of snapshots.
-
Enhancement #2395: Ignore sync errors when operation not supported by local filesystem
The local backend has been modified to work with filesystems which doesn't support the
sync
operation. This operation is normally used by restic to ensure that data files are fully written to disk before continuing.For these limited filesystems, saving a file in the backend would previously fail with an "operation not supported" error. This error is now ignored, which means that e.g. an SMB mount on macOS can now be used as storage location for a repository.
restic#2395 https://forum.restic.net/t/sync-errors-on-mac-over-smb/1859
-
Enhancement #2423: Support user@domain parsing as user
Added the ability for user@domain-like users to be authenticated over SFTP servers.
-
Enhancement #2427: Add flag
--iexclude-file
to backup commandThe backup command now supports the flag
--iexclude-file
which is a case-insensitive version of--exclude-file
. -
Enhancement #2569: Support excluding files by their size
The
backup
command now supports the--exclude-larger-than
option to exclude files which are larger than the specified maximum size. This can for example be useful to exclude unimportant files with a large file size. -
Enhancement #2571: Self-heal missing file parts during backup of unchanged files
We've improved the resilience of restic to certain types of repository corruption.
For files that are unchanged since the parent snapshot, the backup command now verifies that all parts of the files still exist in the repository. Parts that are missing, e.g. from a damaged repository, are backed up again. This verification was already run for files that were modified since the parent snapshot, but is now also done for unchanged files.
Note that restic will not backup file parts that are referenced in the index but where the actual data is not present on disk, as this situation can only be detected by restic check. Please ensure that you run
restic check
regularly. -
Enhancement #2576: Improve the chunking algorithm
We've updated the chunker library responsible for splitting files into smaller blocks. It should improve the chunking throughput by 5-15% depending on the CPU.
-
Enhancement #2598: Improve speed of diff command
We've improved the performance of the diff command when comparing snapshots with similar content. It should run up to twice as fast as before.
-
Enhancement #2599: Slightly reduce memory usage of prune and stats commands
The prune and the stats command kept directory identifiers in memory twice while searching for used blobs.
-
Enhancement #2733: S3 backend: Add support for WebIdentityTokenFile
We've added support for EKS IAM roles for service accounts feature to the S3 backend.
-
Enhancement #2773: Optimize handling of new index entries
Restic now uses less memory for backups which add a lot of data, e.g. large initial backups. In addition, we've improved the stability in some edge cases.
-
Enhancement #2781: Reduce memory consumption of in-memory index
We've improved how the index is stored in memory. This change can reduce memory usage for large repositories by up to 50% (depending on the operation).
-
Enhancement #2786: Optimize
list blobs
commandWe've changed the implementation of
list blobs
which should be now a bit faster and consume almost no memory even for large repositories. -
Enhancement #2790: Optimized file access in restic mount
Reading large (> 100GiB) files from restic mountpoints is now faster, and the speedup is greater for larger files.
-
Enhancement #2840: Speed-up file deletion in forget, prune and rebuild-index
We've sped up the file deletion for the commands forget, prune and rebuild-index, especially for remote repositories. Deletion was sequential before and is now run in parallel.
-
Enhancement #2858: Support filtering snapshots by tag and path in the stats command
We've added filtering snapshots by
--tag tagList
and by--path path
to thestats
command. This includes filtering of only 'latest' snapshots or all snapshots in a repository.restic#2858 restic#2859 https://forum.restic.net/t/stats-for-a-host-and-filtered-snapshots/3020
The following sections list the changes in restic 0.9.6 relevant to restic users. The changes are ordered by importance.
- Fix #2063: Allow absolute path for filename when backing up from stdin
- Fix #2174: Save files with invalid timestamps
- Fix #2249: Read fresh metadata for unmodified files
- Fix #2301: Add upper bound for t in --read-data-subset=n/t
- Fix #2321: Check errors when loading index files
- Enh #2179: Use ctime when checking for file changes
- Enh #2306: Allow multiple retries for interactive password input
- Enh #2330: Make
--group-by
accept both singular and plural - Enh #2350: Add option to configure S3 region
-
Bugfix #2063: Allow absolute path for filename when backing up from stdin
When backing up from stdin, handle directory path for
--stdin-filename
. This can be used to specify the full path for the backed-up file. -
Bugfix #2174: Save files with invalid timestamps
When restic reads invalid timestamps (year is before 0000 or after 9999) it refused to read and archive the file. We've changed the behavior and will now save modified timestamps with the year set to either 0000 or 9999, the rest of the timestamp stays the same, so the file will be saved (albeit with a bogus timestamp).
-
Bugfix #2249: Read fresh metadata for unmodified files
Restic took all metadata for files which were detected as unmodified, not taking into account changed metadata (ownership, mode). This is now corrected.
-
Bugfix #2301: Add upper bound for t in --read-data-subset=n/t
256 is the effective maximum for t, but restic would allow larger values, leading to strange behavior.
-
Bugfix #2321: Check errors when loading index files
Restic now checks and handles errors which occur when loading index files, the missing check leads to odd errors (and a stack trace printed to users) later. This was reported in the forum.
restic#2321 https://forum.restic.net/t/check-rebuild-index-prune/1848/13
-
Enhancement #2179: Use ctime when checking for file changes
Previously, restic only checked a file's mtime (along with other non-timestamp metadata) to decide if a file has changed. This could cause restic to not notice that a file has changed (and therefore continue to store the old version, as opposed to the modified version) if something edits the file and then resets the timestamp. Restic now also checks the ctime of files, so any modifications to a file should be noticed, and the modified file will be backed up. The ctime check will be disabled if the --ignore-inode flag was given.
If this change causes problems for you, please open an issue, and we can look in to adding a seperate flag to disable just the ctime check.
-
Enhancement #2306: Allow multiple retries for interactive password input
Restic used to quit if the repository password was typed incorrectly once. Restic will now ask the user again for the repository password if typed incorrectly. The user will now get three tries to input the correct password before restic quits.
-
Enhancement #2330: Make
--group-by
accept both singular and pluralOne can now use the values
host
/hosts
,path
/paths
andtag
/tags
interchangeably in the--group-by
argument. -
Enhancement #2350: Add option to configure S3 region
We've added a new option for setting the region when accessing an S3-compatible service. For some providers, it is required to set this to a valid value. You can do that either by setting the environment variable
AWS_DEFAULT_REGION
or using the options3.region
, e.g. like this:-o s3.region="us-east-1"
.
The following sections list the changes in restic 0.9.5 relevant to restic users. The changes are ordered by importance.
- Fix #2135: Return error when no bytes could be read from stdin
- Fix #2181: Don't cancel timeout after 30 seconds for self-update
- Fix #2203: Fix reading passwords from stdin
- Fix #2224: Don't abort the find command when a tree can't be loaded
- Enh #1895: Add case insensitive include & exclude options
- Enh #1937: Support streaming JSON output for backup
- Enh #2037: Add group-by option to snapshots command
- Enh #2124: Ability to dump folders to tar via stdout
- Enh #2139: Return error if no bytes could be read for
backup --stdin
- Enh #2155: Add Openstack application credential auth for Swift
- Enh #2184: Add --json support to forget command
- Enh #2205: Add --ignore-inode option to backup cmd
- Enh #2220: Add config option to set S3 storage class
-
Bugfix #2135: Return error when no bytes could be read from stdin
We assume that users reading backup data from stdin want to know when no data could be read, so now restic returns an error when
backup --stdin
is called but no bytes could be read. Usually, this means that an earlier command in a pipe has failed. The documentation was amended and now recommends setting thepipefail
option (set -o pipefail
). -
Bugfix #2181: Don't cancel timeout after 30 seconds for self-update
-
Bugfix #2203: Fix reading passwords from stdin
Passwords for the
init
,key add
, andkey passwd
commands can now be read from non-terminal stdin. -
Bugfix #2224: Don't abort the find command when a tree can't be loaded
Change the find command so that missing trees don't result in a crash. Instead, the error is logged to the debug log, and the tree ID is displayed along with the snapshot it belongs to. This makes it possible to recover repositories that are missing trees by forgetting the snapshots they are used in.
-
Enhancement #1895: Add case insensitive include & exclude options
The backup and restore commands now have --iexclude and --iinclude flags as case insensitive variants of --exclude and --include.
-
Enhancement #1937: Support streaming JSON output for backup
We've added support for getting machine-readable status output during backup, just pass the flag
--json
forrestic backup
and restic will output a stream of JSON objects which contain the current progress. -
Enhancement #2037: Add group-by option to snapshots command
We have added an option to group the output of the snapshots command, similar to the output of the forget command. The option has been called "--group-by" and accepts any combination of the values "host", "paths" and "tags", separated by commas. Default behavior (not specifying --group-by) has not been changed. We have added support of the grouping to the JSON output.
-
Enhancement #2124: Ability to dump folders to tar via stdout
We've added the ability to dump whole folders to stdout via the
dump
command. Restic now requires at least Go 1.10 due to a limitation of the standard library for Go <= 1.9. -
Enhancement #2139: Return error if no bytes could be read for
backup --stdin
When restic is used to backup the output of a program, like
mysqldump | restic backup --stdin
, it now returns an error if no bytes could be read at all. This catches the failure case whenmysqldump
failed for some reason and did not output any data to stdout. -
Enhancement #2155: Add Openstack application credential auth for Swift
Since Openstack Queens Identity (auth V3) service supports an application credential auth method. It allows to create a technical account with the limited roles. This commit adds an application credential authentication method for the Swift backend.
-
Enhancement #2184: Add --json support to forget command
The forget command now supports the --json argument, outputting the information about what is (or would-be) kept and removed from the repository.
-
Enhancement #2205: Add --ignore-inode option to backup cmd
This option handles backup of virtual filesystems that do not keep fixed inodes for files, like Fuse-based, pCloud, etc. Ignoring inode changes allows to consider the file as unchanged if last modification date and size are unchanged.
-
Enhancement #2220: Add config option to set S3 storage class
The
s3.storage-class
option can be passed to restic (using-o
) to specify the storage class to be used for S3 objects created by restic.The storage class is passed as-is to S3, so it needs to be understood by the API. On AWS, it can be one of
STANDARD
,STANDARD_IA
,ONEZONE_IA
,INTELLIGENT_TIERING
andREDUCED_REDUNDANCY
. If unspecified, the default storage class is used (STANDARD
on AWS).You can mix storage classes in the same bucket, and the setting isn't stored in the restic repository, so be sure to specify it with each command that writes to S3.
The following sections list the changes in restic 0.9.4 relevant to restic users. The changes are ordered by importance.
- Fix #1989: Google Cloud Storage: Respect bandwidth limit
- Fix #2040: Add host name filter shorthand flag for
stats
command - Fix #2068: Correctly return error loading data
- Fix #2095: Consistently use local time for snapshots times
- Enh #1605: Concurrent restore
- Enh #2017: Mount: Enforce FUSE Unix permissions with allow-other
- Enh #2070: Make all commands display timestamps in local time
- Enh #2085: Allow --files-from to be specified multiple times
- Enh #2089: Increase granularity of the "keep within" retention policy
- Enh #2094: Run command to get password
- Enh #2097: Add key hinting
-
Bugfix #1989: Google Cloud Storage: Respect bandwidth limit
The GCS backend did not respect the bandwidth limit configured, a previous commit accidentally removed support for it.
-
Bugfix #2040: Add host name filter shorthand flag for
stats
commandThe default value for
--host
flag was set to 'H' (the shorthand version of the flag), this caused the lookup for the latest snapshot to fail.Add shorthand flag
-H
for--host
(with empty default so if these flags are not specified the latest snapshot will not filter by host name).Also add shorthand
-H
forbackup
command. -
Bugfix #2068: Correctly return error loading data
In one case during
prune
andcheck
, an error loading data from the backend is not returned properly. This is now corrected. -
Bugfix #2095: Consistently use local time for snapshots times
By default snapshots created with restic backup were set to local time, but when the --time flag was used the provided timestamp was parsed as UTC. With this change all snapshots times are set to local time.
-
Enhancement #1605: Concurrent restore
This change significantly improves restore performance, especially when using high-latency remote repositories like B2.
The implementation now uses several concurrent threads to download and process multiple remote files concurrently. To further reduce restore time, each remote file is downloaded using a single repository request.
-
Enhancement #2017: Mount: Enforce FUSE Unix permissions with allow-other
The fuse mount (
restic mount
) now lets the kernel check the permissions of the files within snapshots (this is done through theDefaultPermissions
FUSE option) when the option--allow-other
is specified.To restore the old behavior, we've added the
--no-default-permissions
option. This allows all users that have access to the mount point to access all files within the snapshots. -
Enhancement #2070: Make all commands display timestamps in local time
Restic used to drop the timezone information from displayed timestamps, it now converts timestamps to local time before printing them so the times can be easily compared to.
-
Enhancement #2085: Allow --files-from to be specified multiple times
Before, restic took only the last file specified with
--files-from
into account, this is now corrected. -
Enhancement #2089: Increase granularity of the "keep within" retention policy
The
keep-within
option of theforget
command now accepts time ranges with an hourly granularity. For example, runningrestic forget --keep-within 3d12h
will keep all the snapshots made within three days and twelve hours from the time of the latest snapshot. -
Enhancement #2094: Run command to get password
We've added the
--password-command
option which allows specifying a command that restic runs every time the password for the repository is needed, so it can be integrated with a password manager or keyring. The option can also be set via the environment variable$RESTIC_PASSWORD_COMMAND
. -
Enhancement #2097: Add key hinting
Added a new option
--key-hint
and corresponding environment variableRESTIC_KEY_HINT
. The key hint is a key ID to try decrypting first, before other keys in the repository.This change will benefit repositories with many keys; if the correct key hint is supplied then restic only needs to check one key. If the key hint is incorrect (the key does not exist, or the password is incorrect) then restic will check all keys, as usual.
The following sections list the changes in restic 0.9.3 relevant to restic users. The changes are ordered by importance.
- Fix #1935: Remove truncated files from cache
- Fix #1978: Do not return an error when the scanner is slower than backup
- Enh #1766: Restore: suppress lchown errors when not running as root
- Enh #1777: Improve the
find
command - Enh #1876: Display reason why forget keeps snapshots
- Enh #1891: Accept glob in paths loaded via --files-from
- Enh #1909: Reject files/dirs by name first
- Enh #1920: Vendor dependencies with Go 1.11 Modules
- Enh #1940: Add directory filter to ls command
- Enh #1949: Add new command
self-update
- Enh #1953: Ls: Add JSON output support for restic ls cmd
- Enh #1962: Stream JSON output for ls command
- Enh #1967: Use
--host
everywhere - Enh #2028: Display size of cache directories
-
Bugfix #1935: Remove truncated files from cache
When a file in the local cache is truncated, and restic tries to access data beyond the end of the (cached) file, it used to return an error "EOF". This is now fixed, such truncated files are removed and the data is fetched directly from the backend.
-
Bugfix #1978: Do not return an error when the scanner is slower than backup
When restic makes a backup, there's a background task called "scanner" which collects information on how many files and directories are to be saved, in order to display progress information to the user. When the backup finishes faster than the scanner, it is aborted because the result is not needed any more. This logic contained a bug, where quitting the scanner process was treated as an error, and caused restic to print an unhelpful error message ("context canceled").
-
Enhancement #1766: Restore: suppress lchown errors when not running as root
Like "cp" and "rsync" do, restic now only reports errors for changing the ownership of files during restore if it is run as root, on non-Windows operating systems. On Windows, the error is reported as usual.
-
Enhancement #1777: Improve the
find
commandWe've updated the
find
command to support multiple patterns.restic find
is now able to list the snapshots containing a specific tree or blob, or even the snapshots that contain blobs belonging to a given pack. A list of IDs can be given, as long as they all have the same type.The command
find
can also display the pack IDs the blobs belong to, if the--show-pack-id
flag is provided. -
Enhancement #1876: Display reason why forget keeps snapshots
We've added a column to the list of snapshots
forget
keeps which details the reasons to keep a particuliar snapshot. This makes debugging policies for forget much easier. Please remember to always try things out with--dry-run
! -
Enhancement #1891: Accept glob in paths loaded via --files-from
Before that, behaviour was different if paths were appended to command line or from a file, because wild card characters were expanded by shell if appended to command line, but not expanded if loaded from file.
-
Enhancement #1909: Reject files/dirs by name first
The current scanner/archiver code had an architectural limitation: it always ran the
lstat()
system call on all files and directories before a decision to include/exclude the file/dir was made. This lead to a lot of unnecessary system calls for items that could have been rejected by their name or path only.We've changed the archiver/scanner implementation so that it now first rejects by name/path, and only runs the system call on the remaining items. This reduces the number of
lstat()
system calls a lot (depending on the exclude settings). -
Enhancement #1920: Vendor dependencies with Go 1.11 Modules
Until now, we've used
dep
for managing dependencies, we've now switch to using Go modules. For users this does not change much, only if you want to compile restic without downloading anything with Go 1.11, then you need to run:go build -mod=vendor build.go
-
Enhancement #1940: Add directory filter to ls command
The ls command can now be filtered by directories, so that only files in the given directories will be shown. If the --recursive flag is specified, then ls will traverse subfolders and list their files as well.
It used to be possible to specify multiple snapshots, but that has been replaced by only one snapshot and the possibility of specifying multiple directories.
Specifying directories constrains the walk, which can significantly speed up the listing.
-
Enhancement #1949: Add new command
self-update
We have added a new command called
self-update
which downloads the latest released version of restic from GitHub and replaces the current binary with it. It does not rely on any external program (so it'll work everywhere), but still verifies the GPG signature using the embedded GPG public key.By default, the
self-update
command is hidden behind theselfupdate
built tag, which is only set when restic is built usingbuild.go
(including official releases). The reason for this is that downstream distributions will then not include the command by default, so users are encouraged to use the platform-specific distribution mechanism. -
Enhancement #1953: Ls: Add JSON output support for restic ls cmd
We've implemented listing files in the repository with JSON as output, just pass
--json
as an option torestic ls
. This makes the output of the command machine readable. -
Enhancement #1962: Stream JSON output for ls command
The
ls
command now supports JSON output with the global--json
flag, and this change streams out JSON messages one object at a time rather than en entire array buffered in memory before encoding. The advantage is it allows large listings to be handled efficiently.Two message types are printed: snapshots and nodes. A snapshot object will precede node objects which belong to that snapshot. The
struct_type
field can be used to determine which kind of message an object is. -
Enhancement #1967: Use
--host
everywhereWe now use the flag
--host
for all commands which need a host name, using--hostname
(e.g. forrestic backup
) still works, but will print a deprecation warning. Also, add the short option-H
where possible. -
Enhancement #2028: Display size of cache directories
The
cache
command now by default shows the size of the individual cache directories. It can be disabled with--no-size
.
The following sections list the changes in restic 0.9.2 relevant to restic users. The changes are ordered by importance.
- Fix #1854: Allow saving files/dirs on different fs with
--one-file-system
- Fix #1861: Fix case-insensitive search with restic find
- Fix #1870: Fix restore with --include
- Fix #1880: Use
--cache-dir
argument forcheck
command - Fix #1893: Return error when exclude file cannot be read
- Enh #874: Add stats command to get information about a repository
- Enh #1477: S3 backend: accept AWS_SESSION_TOKEN
- Enh #1772: Add restore --verify to verify restored file content
- Enh #1853: Add JSON output support to
restic key list
- Enh #1901: Update the Backblaze B2 library
- Enh #1906: Add support for B2 application keys
-
Bugfix #1854: Allow saving files/dirs on different fs with
--one-file-system
Restic now allows saving files/dirs on a different file system in a subdir correctly even when
--one-file-system
is specified.The first thing the restic archiver code does is to build a tree of the target files/directories. If it detects that a parent directory is already included (e.g.
restic backup /foo /foo/bar/baz
), it'll ignore the latter argument.Without
--one-file-system
, that's perfectly valid: If/foo
is to be archived, it will include/foo/bar/baz
. But with--one-file-system
,/foo/bar/baz
may reside on a different file system, so it won't be included with/foo
. -
Bugfix #1861: Fix case-insensitive search with restic find
We've fixed the behavior for
restic find -i PATTERN
, which was broken in v0.9.1. -
Bugfix #1870: Fix restore with --include
We fixed a bug which prevented restic to restore files with an include filter.
-
Bugfix #1880: Use
--cache-dir
argument forcheck
commandcheck
command now uses a temporary sub-directory of the specified directory if set using the--cache-dir
argument. If not set, the cache directory is created in the default temporary directory as before. In either case a temporary cache is used to ensure the actual repository is checked (rather than a local copy).The
--cache-dir
argument was not used by thecheck
command, instead a cache directory was created in the temporary directory. -
Bugfix #1893: Return error when exclude file cannot be read
A bug was found: when multiple exclude files were passed to restic and one of them could not be read, an error was printed and restic continued, ignoring even the existing exclude files. Now, an error message is printed and restic aborts when an exclude file cannot be read.
-
Enhancement #874: Add stats command to get information about a repository
-
Enhancement #1477: S3 backend: accept AWS_SESSION_TOKEN
Before, it was not possible to use s3 backend with AWS temporary security credentials(with AWS_SESSION_TOKEN). This change gives higher priority to credentials.EnvAWS credentials provider.
-
Enhancement #1772: Add restore --verify to verify restored file content
Restore will print error message if restored file content does not match expected SHA256 checksum
-
Enhancement #1853: Add JSON output support to
restic key list
This PR enables users to get the output of
restic key list
in JSON in addition to the existing table format. -
Enhancement #1901: Update the Backblaze B2 library
We've updated the library we're using for accessing the Backblaze B2 service to 0.5.0 to include support for upcoming so-called "application keys". With this feature, you can create access credentials for B2 which are restricted to e.g. a single bucket or even a sub-directory of a bucket.
-
Enhancement #1906: Add support for B2 application keys
Restic can now use so-called "application keys" which can be created in the B2 dashboard and were only introduced recently. In contrast to the "master key", such keys can be restricted to a specific bucket and/or path.
The following sections list the changes in restic 0.9.1 relevant to restic users. The changes are ordered by importance.
- Fix #1801: Add limiting bandwidth to the rclone backend
- Fix #1822: Allow uploading large files to MS Azure
- Fix #1825: Correct
find
to not skip snapshots - Fix #1833: Fix caching files on error
- Fix #1834: Resolve deadlock
-
Bugfix #1801: Add limiting bandwidth to the rclone backend
The rclone backend did not respect
--limit-upload
or--limit-download
. Oftentimes it's not necessary to use this, as the limiting in rclone itself should be used because it gives much better results, but in case a remote instance of rclone is used (e.g. called via ssh), it is still relevant to limit the bandwidth from restic to rclone. -
Bugfix #1822: Allow uploading large files to MS Azure
Sometimes, restic creates files to be uploaded to the repository which are quite large, e.g. when saving directories with many entries or very large files. The MS Azure API does not allow uploading files larger that 256MiB directly, rather restic needs to upload them in blocks of 100MiB. This is now implemented.
-
Bugfix #1825: Correct
find
to not skip snapshotsUnder certain circumstances, the
find
command was found to skip snapshots containing directories with files to look for when the directories haven't been modified at all, and were already printed as part of a different snapshot. This is now corrected.In addition, we've switched to our own matching/pattern implementation, so now things like
restic find "/home/user/foo/**/main.go"
are possible. -
Bugfix #1833: Fix caching files on error
During
check
it may happen that different threads access the same file in the backend, which is then downloaded into the cache only once. When that fails, only the thread which is responsible for downloading the file signals the correct error. The other threads just assume that the file has been downloaded successfully and then get an error when they try to access the cached file. -
Bugfix #1834: Resolve deadlock
When the "scanning" process restic runs to find out how much data there is does not finish before the backup itself is done, restic stops doing anything. This is resolved now.
The following sections list the changes in restic 0.9.0 relevant to restic users. The changes are ordered by importance.
- Fix #1608: Respect time stamp for new backup when reading from stdin
- Fix #1652: Ignore/remove invalid lock files
- Fix #1684: Fix backend tests for rest-server
- Fix #1730: Ignore sockets for restore
- Fix #1745: Correctly parse the argument to --tls-client-cert
- Enh #549: Rework archiver code
- Enh #827: Add --new-password-file flag for non-interactive password changes
- Enh #1433: Support UTF-16 encoding and process Byte Order Mark
- Enh #1477: Accept AWS_SESSION_TOKEN for the s3 backend
- Enh #1552: Use Google Application Default credentials
- Enh #1561: Allow using rclone to access other services
- Enh #1648: Ignore AWS permission denied error when creating a repository
- Enh #1649: Add illumos/Solaris support
- Enh #1665: Improve cache handling for
restic check
- Enh #1709: Improve messages
restic check
prints - Enh #1721: Add
cache
command to list cache dirs - Enh #1735: Allow keeping a time range of snaphots
- Enh #1758: Allow saving OneDrive folders in Windows
- Enh #1782: Use default AWS credentials chain for S3 backend
-
Bugfix #1608: Respect time stamp for new backup when reading from stdin
When reading backups from stdin (via
restic backup --stdin
), restic now uses the time stamp for the new backup passed in--time
. -
Bugfix #1652: Ignore/remove invalid lock files
This corrects a bug introduced recently: When an invalid lock file in the repo is encountered (e.g. if the file is empty), the code used to ignore that, but now returns the error. Now, invalid files are ignored for the normal lock check, and removed when
restic unlock --remove-all
is run. -
Bugfix #1684: Fix backend tests for rest-server
The REST server for restic now requires an explicit parameter (
--no-auth
) if no authentication should be allowed. This is fixed in the tests. -
Bugfix #1730: Ignore sockets for restore
We've received a report and correct the behavior in which the restore code aborted restoring a directory when a socket was encountered. Unix domain socket files cannot be restored (they are created on the fly once a process starts listening). The error handling was corrected, and in addition we're now ignoring sockets during restore.
-
Bugfix #1745: Correctly parse the argument to --tls-client-cert
Previously, the --tls-client-cert method attempt to read ARGV[1] (hardcoded) instead of the argument that was passed to it. This has been corrected.
-
Enhancement #549: Rework archiver code
The core archiver code and the complementary code for the
backup
command was rewritten completely. This resolves very annoying issues such as 549. The first backup with this release of restic will likely result in all files being re-read locally, so it will take a lot longer. The next backup after that will be fast again.Basically, with the old code, restic took the last path component of each to-be-saved file or directory as the top-level file/directory within the snapshot. This meant that when called as
restic backup /home/user/foo
, the snapshot would contain the files in the directory/home/user/foo
as/foo
.This is not the case any more with the new archiver code. Now, restic works very similar to what
tar
does: When restic is called with an absolute path to save, then it'll preserve the directory structure within the snapshot. For the example above, the snapshot would contain the files in the directory within/home/user/foo
in the snapshot. For relative directories, it only preserves the relative path components. Sorestic backup user/foo
will save the files as/user/foo
in the snapshot.While we were at it, the status display and notification system was completely rewritten. By default, restic now shows which files are currently read (unless
--quiet
is specified) in a multi-line status display.The
backup
command also gained a new option:--verbose
. It can be specified once (which prints a bit more detail what restic is doing) or twice (which prints a line for each file/directory restic encountered, together with some statistics).Another issue that was resolved is the new code only reads two files at most. The old code would read way too many files in parallel, thereby slowing down the backup process on spinning discs a lot.
restic#549 restic#1286 restic#446 restic#1344 restic#1416 restic#1456 restic#1145 restic#1160 restic#1494
-
Enhancement #827: Add --new-password-file flag for non-interactive password changes
This makes it possible to change a repository password without being prompted.
restic#827 restic#1720 https://forum.restic.net/t/changing-repo-password-without-prompt/591
-
Enhancement #1433: Support UTF-16 encoding and process Byte Order Mark
On Windows, text editors commonly leave a Byte Order Mark at the beginning of the file to define which encoding is used (oftentimes UTF-16). We've added code to support processing the BOMs in text files, like the exclude files, the password file and the file passed via
--files-from
. This does not apply to any file being saved in a backup, those are not touched and archived as they are. -
Enhancement #1477: Accept AWS_SESSION_TOKEN for the s3 backend
Before, it was not possible to use s3 backend with AWS temporary security credentials(with AWS_SESSION_TOKEN). This change gives higher priority to credentials.EnvAWS credentials provider.
-
Enhancement #1552: Use Google Application Default credentials
Google provide libraries to generate appropriate credentials with various fallback sources. This change uses the library to generate our GCS client, which allows us to make use of these extra methods.
This should be backward compatible with previous restic behaviour while adding the additional capabilities to auth from Google's internal metadata endpoints. For users running restic in GCP this can make authentication far easier than it was before.
restic#1552 https://developers.google.com/identity/protocols/application-default-credentials
-
Enhancement #1561: Allow using rclone to access other services
We've added the ability to use rclone to store backup data on all backends that it supports. This was done in collaboration with Nick, the author of rclone. You can now use it to first configure a service, then restic manages the rest (starting and stopping rclone). For details, please see the manual.
-
Enhancement #1648: Ignore AWS permission denied error when creating a repository
It's not possible to use s3 backend scoped to a subdirectory(with specific permissions). Restic doesn't try to create repository in a subdirectory, when 'bucket exists' of parent directory check fails due to permission issues.
-
Enhancement #1649: Add illumos/Solaris support
-
Enhancement #1665: Improve cache handling for
restic check
For safety reasons, restic does not use a local metadata cache for the
restic check
command, so that data is loaded from the repository and restic can check it's in good condition. When the cache is disabled, restic will fetch each tiny blob needed for checking the integrity using a separate backend request. For non-local backends, that will take a long time, and depending on the backend (e.g. B2) may also be much more expensive.This PR adds a few commits which will change the behavior as follows:
-
When
restic check
is called without any additional parameters, it will build a new cache in a temporary directory, which is removed at the end of the check. This way, we'll get readahead for metadata files (so restic will fetch the whole file when the first blob from the file is requested), but all data is freshly fetched from the storage backend. This is the default behavior and will work for almost all users. -
When
restic check
is called with--with-cache
, the default on-disc cache is used. This behavior hasn't changed since the cache was introduced. -
When
--no-cache
is specified, restic falls back to the old behavior, and read all tiny blobs in separate requests.
-
-
Enhancement #1709: Improve messages
restic check
printsSome messages
restic check
prints are not really errors, so from now on restic does not treat them as errors any more and exits cleanly.restic#1709 https://forum.restic.net/t/what-is-the-standard-procedure-to-follow-if-a-backup-or-restore-is-interrupted/571/2
-
Enhancement #1721: Add
cache
command to list cache dirsThe command
cache
was added, it allows listing restic's cache directoriers together with the last usage. It also allows removing old cache dirs without having to access a repo, viarestic cache --cleanup
-
Enhancement #1735: Allow keeping a time range of snaphots
We've added the
--keep-within
option to theforget
command. It instructs restic to keep all snapshots within the given duration since the newest snapshot. For example, runningrestic forget --keep-within 5m7d
will keep all snapshots which have been made in the five months and seven days since the latest snapshot. -
Enhancement #1758: Allow saving OneDrive folders in Windows
Restic now contains a bugfix to two libraries, which allows saving OneDrive folders in Windows. In order to use the newer versions of the libraries, the minimal version required to compile restic is now Go 1.9.
-
Enhancement #1782: Use default AWS credentials chain for S3 backend
Adds support for file credentials to the S3 backend (e.g. ~/.aws/credentials), and reorders the credentials chain for the S3 backend to match AWS's standard, which is static credentials, env vars, credentials file, and finally remote.
The following sections list the changes in restic 0.8.3 relevant to restic users. The changes are ordered by importance.
- Fix #1633: Fixed unexpected 'pack file cannot be listed' error
- Fix #1638: Handle errors listing files in the backend
- Fix #1641: Ignore files with invalid names in the repo
- Enh #1497: Add --read-data-subset flag to check command
- Enh #1560: Retry all repository file download errors
- Enh #1623: Don't check for presence of files in the backend before writing
- Enh #1634: Upgrade B2 client library, reduce HTTP requests
-
Bugfix #1633: Fixed unexpected 'pack file cannot be listed' error
Due to a regression introduced in 0.8.2, the
rebuild-index
andprune
commands failed to read pack files with size of 587, 588, 589 or 590 bytes. -
Bugfix #1638: Handle errors listing files in the backend
A user reported in the forum that restic completes a backup although a concurrent
prune
operation was running. A few error messages were printed, but the backup was attempted and completed successfully. No error code was returned.This should not happen: The repository is exclusively locked during
prune
, so whenrestic backup
is run in parallel, it should abort and return an error code instead.It was found that the bug was in the code introduced only recently, which retries a List() operation on the backend should that fail. It is now corrected.
restic#1638 https://forum.restic.net/t/restic-backup-returns-0-exit-code-when-already-locked/484
-
Bugfix #1641: Ignore files with invalid names in the repo
The release 0.8.2 introduced a bug: when restic encounters files in the repo which do not have a valid name, it tries to load a file with a name of lots of zeroes instead of ignoring it. This is now resolved, invalid file names are just ignored.
restic#1641 restic#1643 https://forum.restic.net/t/help-fixing-repo-no-such-file/485/3
-
Enhancement #1497: Add --read-data-subset flag to check command
This change introduces ability to check integrity of a subset of repository data packs. This can be used to spread integrity check of larger repositories over a period of time.
-
Enhancement #1560: Retry all repository file download errors
Restic will now retry failed downloads, similar to other operations.
-
Enhancement #1623: Don't check for presence of files in the backend before writing
Before, all backend implementations were required to return an error if the file that is to be written already exists in the backend. For most backends, that means making a request (e.g. via HTTP) and returning an error when the file already exists.
This is not accurate, the file could have been created between the HTTP request testing for it, and when writing starts, so we've relaxed this requeriment, which saves one additional HTTP request per newly added file.
-
Enhancement #1634: Upgrade B2 client library, reduce HTTP requests
We've upgraded the B2 client library restic uses to access BackBlaze B2. This reduces the number of HTTP requests needed to upload a new file from two to one, which should improve throughput to B2.
The following sections list the changes in restic 0.8.2 relevant to restic users. The changes are ordered by importance.
- Fix #1506: Limit bandwith at the http.RoundTripper for HTTP based backends
- Fix #1512: Restore directory permissions as the last step
- Fix #1528: Correctly create missing subdirs in data/
- Fix #1589: Complete intermediate index upload
- Fix #1590: Strip spaces for lines read via --files-from
- Fix #1594: Google Cloud Storage: Use generic HTTP transport
- Fix #1595: Backup: Remove bandwidth display
- Enh #1507: Only reload snapshots once per minute for fuse mount
- Enh #1522: Add support for TLS client certificate authentication
- Enh #1538: Reduce memory allocations for querying the index
- Enh #1541: Reduce number of remote requests during repository check
- Enh #1549: Speed up querying across indices and scanning existing files
- Enh #1554: Fuse/mount: Correctly handle EOF, add template option
- Enh #1564: Don't terminate ssh on SIGINT
- Enh #1567: Reduce number of backend requests for rebuild-index and prune
- Enh #1579: Retry Backend.List() in case of errors
- Enh #1584: Limit index file size
-
Bugfix #1506: Limit bandwith at the http.RoundTripper for HTTP based backends
-
Bugfix #1512: Restore directory permissions as the last step
This change allows restoring into directories that were not writable during backup. Before, restic created the directory, set the read-only mode and then failed to create files in the directory. This change now restores the directory (with its permissions) as the very last step.
-
Bugfix #1528: Correctly create missing subdirs in data/
-
Bugfix #1589: Complete intermediate index upload
After a user posted a comprehensive report of what he observed, we were able to find a bug and correct it: During backup, restic uploads so-called "intermediate" index files. When the backup finishes during a transfer of such an intermediate index, the upload is cancelled, but the backup is finished without an error. This leads to an inconsistent state, where the snapshot references data that is contained in the repo, but is not referenced in any index.
The situation can be resolved by building a new index with
rebuild-index
, but looks very confusing at first. Since all the data got uploaded to the repo successfully, there was no risk of data loss, just minor inconvenience for our users.restic#1589 https://forum.restic.net/t/error-loading-tree-check-prune-and-forget-gives-error-b2-backend/406
-
Bugfix #1590: Strip spaces for lines read via --files-from
Leading and trailing spaces in lines read via
--files-from
are now stripped, so it behaves the same as with lines read via--exclude-file
. -
Bugfix #1594: Google Cloud Storage: Use generic HTTP transport
It was discovered that the Google Cloud Storage backend did not use the generic HTTP transport, so things such as bandwidth limiting with
--limit-upload
did not work. This is resolved now. -
Bugfix #1595: Backup: Remove bandwidth display
This commit removes the bandwidth displayed during backup process. It is misleading and seldomly correct, because it's neither the "read bandwidth" (only for the very first backup) nor the "upload bandwidth". Many users are confused about (and rightly so), c.f. #1581, #1033, #1591
We'll eventually replace this display with something more relevant when the new archiver code is ready.
-
Enhancement #1507: Only reload snapshots once per minute for fuse mount
-
Enhancement #1522: Add support for TLS client certificate authentication
Support has been added for using a TLS client certificate for authentication to HTTP based backend. A file containing the PEM encoded private key and certificate can be set using the
--tls-client-cert
option. -
Enhancement #1538: Reduce memory allocations for querying the index
This change reduces the internal memory allocations when the index data structures in memory are queried if a blob (part of a file) already exists in the repo. It should speed up backup a bit, and maybe even reduce RAM usage.
-
Enhancement #1541: Reduce number of remote requests during repository check
This change eliminates redundant remote repository calls and significantly improves repository check time.
-
Enhancement #1549: Speed up querying across indices and scanning existing files
This change increases the whenever a blob (part of a file) is searched for in a restic repository. This will reduce cpu usage some when backing up files already backed up by restic. Cpu usage is further decreased when scanning files.
-
Enhancement #1554: Fuse/mount: Correctly handle EOF, add template option
We've added the
--snapshot-template
string, which can be used to specify a template for a snapshot directory. In addition, accessing data after the end of a file via the fuse mount is now handled correctly. -
Enhancement #1564: Don't terminate ssh on SIGINT
We've reworked the code which runs the
ssh
login for the sftp backend so that it can prompt for a password (if needed) but does not exit when the user presses CTRL+C (SIGINT) e.g. during backup. This allows restic to properly shut down when it receives SIGINT and remove the lock file from the repo, afterwards exiting thessh
process. -
Enhancement #1567: Reduce number of backend requests for rebuild-index and prune
We've found a way to reduce then number of backend requests for the
rebuild-index
andprune
operations. This significantly speeds up the operations for high-latency backends. -
Enhancement #1579: Retry Backend.List() in case of errors
-
Enhancement #1584: Limit index file size
Before, restic would create a single new index file on
prune
orrebuild-index
, this may lead to memory problems when this huge index is created and loaded again. We're now limiting the size of the index file, and split newly created index files into several smaller ones. This allows restic to be more memory-efficient.
The following sections list the changes in restic 0.8.1 relevant to restic users. The changes are ordered by importance.
- Fix #1454: Correct cache dir location for Windows and Darwin
- Fix #1457: Improve s3 backend with DigitalOcean Spaces
- Fix #1459: Disable handling SIGPIPE
- Chg #1452: Do not save atime by default
- Enh #11: Add the
diff
command - Enh #1436: Add code to detect old cache directories
- Enh #1439: Improve cancellation logic
-
Bugfix #1454: Correct cache dir location for Windows and Darwin
The cache directory on Windows and Darwin was not correct, instead the directory
.cache
was used. -
Bugfix #1457: Improve s3 backend with DigitalOcean Spaces
-
Bugfix #1459: Disable handling SIGPIPE
We've disabled handling SIGPIPE again. Turns out, writing to broken TCP connections also raised SIGPIPE, so restic exits on the first write to a broken connection. Instead, restic should retry the request.
-
Change #1452: Do not save atime by default
By default, the access time for files and dirs is not saved any more. It is not possible to reliably disable updating the access time during a backup, so for the next backup the access time is different again. This means a lot of metadata is saved. If you want to save the access time anyway, pass
--with-atime
to thebackup
command. -
Enhancement #11: Add the
diff
commandThe command
diff
was added, it allows comparing two snapshots and listing all differences. -
Enhancement #1436: Add code to detect old cache directories
We've added code to detect old cache directories of repositories that haven't been used in a long time, restic now prints a note when it detects that such dirs exist. Also, the option
--cleanup-cache
was added to automatically remove such directories. That's not a problem because the cache will be rebuild once a repo is accessed again. -
Enhancement #1439: Improve cancellation logic
The cancellation logic was improved, restic can now shut down cleanly when requested to do so (e.g. via ctrl+c).
The following sections list the changes in restic 0.8.0 relevant to restic users. The changes are ordered by importance.
- Sec #1445: Prevent writing outside the target directory during restore
- Fix #1256: Re-enable workaround for S3 backend
- Fix #1291: Reuse backend TCP connections to BackBlaze B2
- Fix #1317: Run prune when
forget --prune
is called with just snapshot IDs - Fix #1437: Remove implicit path
/restic
for the s3 backend - Enh #448: Sftp backend prompts for password
- Enh #510: Add
dump
command - Enh #1040: Add local metadata cache
- Enh #1102: Add subdirectory
ids
to fuse mount - Enh #1114: Add
--cacert
to specify TLS certificates to check against - Enh #1216: Add upload/download limiting
- Enh #1249: Add
latest
symlink in fuse mount - Enh #1269: Add
--compact
toforget
command - Enh #1271: Cache results for excludes for
backup
- Enh #1274: Add
generate
command, replacesmanpage
andautocomplete
- Enh #1281: Google Cloud Storage backend needs less permissions
- Enh #1319: Make
check
printno errors found
explicitly - Enh #1353: Retry failed backend requests
- Enh #1367: Allow comments in files read from via
--file-from
-
Security #1445: Prevent writing outside the target directory during restore
A vulnerability was found in the restic restorer, which allowed attackers in special circumstances to restore files to a location outside of the target directory. Due to the circumstances we estimate this to be a low-risk vulnerability, but urge all users to upgrade to the latest version of restic.
Exploiting the vulnerability requires a Linux/Unix system which saves backups via restic and a Windows systems which restores files from the repo. In addition, the attackers need to be able to create files with arbitrary names which are then saved to the restic repo. For example, by creating a file named "..\test.txt" (which is a perfectly legal filename on Linux) and restoring a snapshot containing this file on Windows, it would be written to the parent of the target directory.
We'd like to thank Tyler Spivey for reporting this responsibly!
-
Bugfix #1256: Re-enable workaround for S3 backend
We've re-enabled a workaround for
minio-go
(the library we're using to access s3 backends), this reduces memory usage. -
Bugfix #1291: Reuse backend TCP connections to BackBlaze B2
A bug was discovered in the library we're using to access Backblaze, it now reuses already established TCP connections which should be a lot faster and not cause network failures any more.
-
Bugfix #1317: Run prune when
forget --prune
is called with just snapshot IDsA bug in the
forget
command causedprune
not to be run when--prune
was specified without a policy, e.g. when only snapshot IDs that should be forgotten are listed manually. -
Bugfix #1437: Remove implicit path
/restic
for the s3 backendThe s3 backend used the subdir
restic
within a bucket if no explicit path after the bucket name was specified. Since this version, restic does not use this default path any more. If you created a repo on s3 in a bucket without specifying a path within the bucket, you need to add/restic
at the end of the repository specification to access your repo:s3:s3.amazonaws.com/bucket/restic
-
Enhancement #448: Sftp backend prompts for password
The sftp backend now prompts for the password if a password is necessary for login.
-
Enhancement #510: Add
dump
commandWe've added the
dump
command which prints a file from a snapshot to stdout. This can e.g. be used to restore files read withbackup --stdin
. -
Enhancement #1040: Add local metadata cache
We've added a local cache for metadata so that restic doesn't need to load all metadata (snapshots, indexes, ...) from the repo each time it starts. By default the cache is active, but there's a new global option
--no-cache
that can be used to disable the cache. By deafult, the cache a standard cache folder for the OS, which can be overridden with--cache-dir
. The cache will automatically populate, indexes and snapshots are saved as they are loaded. Cache directories for repos that haven't been used recently can automatically be removed by restic with the--cleanup-cache
option.A related change was to by default create pack files in the repo that contain either data or metadata, not both mixed together. This allows easy caching of only the metadata files. The next run of
restic prune
will untangle mixed files automatically.restic#29 restic#738 restic#282 restic#1040 restic#1287 restic#1436 restic#1265
-
Enhancement #1102: Add subdirectory
ids
to fuse mountThe fuse mount now has an
ids
subdirectory which contains the snapshots below their (short) IDs. -
Enhancement #1114: Add
--cacert
to specify TLS certificates to check againstWe've added the
--cacert
option which can be used to pass one (or more) CA certificates to restic. These are used in addition to the system CA certificates to verify HTTPS certificates (e.g. for the REST backend). -
Enhancement #1216: Add upload/download limiting
We've added support for rate limiting through
--limit-upload
and--limit-download
flags. -
Enhancement #1249: Add
latest
symlink in fuse mountThe directory structure in the fuse mount now exposes a symlink
latest
which points to the latest snapshot in that particular directory. -
Enhancement #1269: Add
--compact
toforget
commandThe option
--compact
was added to theforget
command to provide the same compact view as thesnapshots
command. -
Enhancement #1271: Cache results for excludes for
backup
The
backup
command now caches the result of excludes for a directory. -
Enhancement #1274: Add
generate
command, replacesmanpage
andautocomplete
The
generate
command has been added, which replaces the now removed commandsmanpage
andautocomplete
. This release of restic contains the most recent manpages indoc/man
and the auto-completion files for bash and zsh indoc/bash-completion.sh
anddoc/zsh-completion.zsh
-
Enhancement #1281: Google Cloud Storage backend needs less permissions
The Google Cloud Storage backend no longer requires the service account to have the
storage.buckets.get
permission ("Storage Admin" role) inrestic init
if the bucket already exists. -
Enhancement #1319: Make
check
printno errors found
explicitlyThe
check
command now explicetly printsNo errors were found
when no errors could be found. -
Enhancement #1353: Retry failed backend requests
-
Enhancement #1367: Allow comments in files read from via
--file-from
When the list of files/dirs to be saved is read from a file with
--files-from
, comment lines (starting with#
) are now ignored.
The following sections list the changes in restic 0.7.3 relevant to restic users. The changes are ordered by importance.
- Fix #1246: List all files stored in Google Cloud Storage
-
Bugfix #1246: List all files stored in Google Cloud Storage
For large backups stored in Google Cloud Storage, the
prune
command fails because listing only returns the first 1000 files. This has been corrected, no data is lost in the process. In addition, a plausibility check was added toprune
.
The following sections list the changes in restic 0.7.2 relevant to restic users. The changes are ordered by importance.
- Fix #1164: Make the
key remove
command behave as documented - Fix #1167: Do not create a local repo unless
init
is used - Fix #1191: Make sure to write profiling files on interrupt
- Enh #317: Add
--exclude-caches
and--exclude-if-present
- Enh #697: Automatically generate man pages for all restic commands
- Enh #1044: Improve
restore
, do not traverse/load excluded directories - Enh #1061: Add Dockerfile and official Docker image
- Enh #1126: Use the standard Go git repository layout, use
dep
for vendoring - Enh #1132: Make
key
command always prompt for a password - Enh #1134: Add support for storing backups on Google Cloud Storage
- Enh #1144: Properly report errors when reading files with exclude patterns
- Enh #1149: Add support for storing backups on Microsoft Azure Blob Storage
- Enh #1179: Resolve name conflicts, append a counter
- Enh #1196: Add
--group-by
toforget
command for flexible grouping - Enh #1203: Print stats on all BSD systems when SIGINFO (ctrl+t) is received
- Enh #1205: Allow specifying time/date for a backup with
--time
- Enh #1218: Add
--compact
tosnapshots
command
-
Bugfix #1164: Make the
key remove
command behave as documented -
Bugfix #1167: Do not create a local repo unless
init
is usedWhen a restic command other than
init
is used with a local repository and the repository directory does not exist, restic creates the directory structure. That's an error, only theinit
command should create the dir. -
Bugfix #1191: Make sure to write profiling files on interrupt
Since a few releases restic had the ability to write profiling files for memory and CPU usage when
debug
is enabled. It was discovered that when restic is interrupted (ctrl+c is pressed), the proper shutdown hook is not run. This is now corrected. -
Enhancement #317: Add
--exclude-caches
and--exclude-if-present
A new option
--exclude-caches
was added that allows excluding cache directories (that are tagged as such). This is a special case of a more generic option--exclude-if-present
which excludes a directory if a file with a specific name (and contents) is present. -
Enhancement #697: Automatically generate man pages for all restic commands
-
Enhancement #1044: Improve
restore
, do not traverse/load excluded directories -
Enhancement #1061: Add Dockerfile and official Docker image
-
Enhancement #1126: Use the standard Go git repository layout, use
dep
for vendoringThe git repository layout was changed to resemble the layout typically used in Go projects, we're not using
gb
for building restic any more and vendoring the dependencies is now taken care of bydep
. -
Enhancement #1132: Make
key
command always prompt for a passwordThe
key
command now prompts for a password even if the original password to access a repo has been specified via theRESTIC_PASSWORD
environment variable or a password file. -
Enhancement #1134: Add support for storing backups on Google Cloud Storage
-
Enhancement #1144: Properly report errors when reading files with exclude patterns
-
Enhancement #1149: Add support for storing backups on Microsoft Azure Blob Storage
The library we're using to access the service requires Go 1.8, so restic now needs at least Go 1.8.
-
Enhancement #1179: Resolve name conflicts, append a counter
-
Enhancement #1196: Add
--group-by
toforget
command for flexible grouping -
Enhancement #1203: Print stats on all BSD systems when SIGINFO (ctrl+t) is received
-
Enhancement #1205: Allow specifying time/date for a backup with
--time
-
Enhancement #1218: Add
--compact
tosnapshots
commandThe option
--compact
was added to thesnapshots
command to get a better overview of the snapshots in a repo. It limits each snapshot to a single line.
The following sections list the changes in restic 0.7.1 relevant to restic users. The changes are ordered by importance.
- Fix #1115: Fix
prune
, only include existing files in indexes - Enh #1055: Create subdirs below
data/
for local/sftp backends - Enh #1067: Allow loading credentials for s3 from IAM
- Enh #1073: Add
migrate
cmd to migrate froms3legacy
todefault
layout - Enh #1080: Ignore chmod() errors on filesystems which do not support it
- Enh #1081: Clarify semantic for
--tag
for theforget
command - Enh #1082: Print stats on SIGINFO on Darwin and FreeBSD (ctrl+t)
-
Bugfix #1115: Fix
prune
, only include existing files in indexesA bug was found (and corrected) in the index rebuilding after prune, which led to indexes which include blobs that were not present in the repo any more. There were already checks in place which detected this situation and aborted with an error message. A new run of either
prune
orrebuild-index
corrected the index files. This is now fixed and a test has been added to detect this. -
Enhancement #1055: Create subdirs below
data/
for local/sftp backendsThe local and sftp backends now create the subdirs below
data/
on open/init. This way, restic makes sure that they always exist. This is connected to an issue for the sftp server.restic#1055 restic/rest-server#11 (comment) restic#1077 restic#1105
-
Enhancement #1067: Allow loading credentials for s3 from IAM
When no S3 credentials are specified in the environment variables, restic now tries to load credentials from an IAM instance profile when the s3 backend is used.
-
Enhancement #1073: Add
migrate
cmd to migrate froms3legacy
todefault
layoutThe
migrate
command for changing thes3legacy
layout to thedefault
layout for s3 backends has been improved: It can now be restarted withrestic migrate --force s3_layout
and automatically retries operations on error. -
Enhancement #1080: Ignore chmod() errors on filesystems which do not support it
-
Enhancement #1081: Clarify semantic for
--tag
for theforget
command -
Enhancement #1082: Print stats on SIGINFO on Darwin and FreeBSD (ctrl+t)
The following sections list the changes in restic 0.7.0 relevant to restic users. The changes are ordered by importance.
- Fix #965: Switch to
default
repo layout for the s3 backend - Fix #1013: Switch back to using the high-level minio-go API for s3
- Enh #512: Add Backblaze B2 backend
- Enh #636: Add dirs
tags
andhosts
to fuse mount - Enh #975: Add new backend for OpenStack Swift
- Enh #989: Improve performance of the
find
command - Enh #998: Improve performance of the fuse mount
- Enh #1021: Detect invalid backend name and print error
- Enh #1029: Remove invalid pack files when
prune
is run
-
Bugfix #965: Switch to
default
repo layout for the s3 backendThe default layout for the s3 backend is now
default
(instead ofs3legacy
). Also, there's a newmigrate
command to convert an existing repo, it can be run like this:restic migrate s3_layout
-
Bugfix #1013: Switch back to using the high-level minio-go API for s3
For the s3 backend we're back to using the high-level API the s3 client library for uploading data, a few users reported dropped connections (which the library will automatically retry now).
-
Enhancement #512: Add Backblaze B2 backend
-
Enhancement #636: Add dirs
tags
andhosts
to fuse mountThe fuse mount now has two more directories:
tags
contains a subdir for each tag, which in turn contains only the snapshots that have this tag. The subdirhosts
contains a subdir for each host that has a snapshot, and the subdir contains the snapshots for that host. -
Enhancement #975: Add new backend for OpenStack Swift
-
Enhancement #989: Improve performance of the
find
commandImproved performance for the
find
command: Restic recognizes paths it has already checked for the files in question, so the number of backend requests is reduced a lot. -
Enhancement #998: Improve performance of the fuse mount
Listing directories which contain large files now is significantly faster.
-
Enhancement #1021: Detect invalid backend name and print error
Restic now tries to detect when an invalid/unknown backend is used and returns an error message.
-
Enhancement #1029: Remove invalid pack files when
prune
is runThe
prune
command has been improved and will now remove invalid pack files, for example files that have not been uploaded completely because a backup was interrupted.
The following sections list the changes in restic 0.6.1 relevant to restic users. The changes are ordered by importance.
- Enh #974: Remove regular status reports
- Enh #981: Remove temporary path from binary in
build.go
- Enh #985: Allow multiple parallel idle HTTP connections
-
Enhancement #974: Remove regular status reports
Regular status report: We've removed the status report that was printed every 10 seconds when restic is run non-interactively. You can still force reporting the current status by sending a
USR1
signal to the process. -
Enhancement #981: Remove temporary path from binary in
build.go
The
build.go
now strips the temporary directory used for compilation from the binary. This is the first step in enabling reproducible builds. -
Enhancement #985: Allow multiple parallel idle HTTP connections
Backends based on HTTP now allow several idle connections in parallel. This is especially important for the REST backend, which (when used with a local server) may create a lot connections and exhaust available ports quickly.
The following sections list the changes in restic 0.6.0 relevant to restic users. The changes are ordered by importance.
- Enh #957: Make
forget
consistent - Enh #962: Improve memory and runtime for the s3 backend
- Enh #966: Unify repository layout for all backends
-
Enhancement #957: Make
forget
consistentThe
forget
command was corrected to be more consistent in which snapshots are to be forgotten. It is possible that the new code removes more snapshots than before, so please review what would be deleted by using the--dry-run
option. -
Enhancement #962: Improve memory and runtime for the s3 backend
We've updated the library used for accessing s3, switched to using a lower level API and added caching for some requests. This lead to a decrease in memory usage and a great speedup. In addition, we added benchmark functions for all backends, so we can track improvements over time. The Continuous Integration test service we're using (Travis) now runs the s3 backend tests not only against a Minio server, but also against the Amazon s3 live service, so we should be notified of any regressions much sooner.
-
Enhancement #966: Unify repository layout for all backends
Up to now the s3 backend used a special repository layout. We've decided to unify the repository layout and implemented the default layout also for the s3 backend. For creating a new repository on s3 with the default layout, use
restic -o s3.layout=default init
. For further commands the option is not necessary any more, restic will automatically detect the correct layout to use. A future version will switch to the default layout for new repositories.