Skip to content

Latest commit

 

History

History
185 lines (138 loc) · 8.43 KB

rootless.md

File metadata and controls

185 lines (138 loc) · 8.43 KB
Raw

Rootless mode

See https://rootlesscontaine.rs/getting-started/common/ for the prerequisites.

Daemon (containerd)

Use containerd-rootless-setuptool.sh to set up rootless containerd.

$ containerd-rootless-setuptool.sh install
[INFO] Checking RootlessKit functionality
[INFO] Checking cgroup v2
[INFO] Checking overlayfs
[INFO] Creating /home/testuser/.config/systemd/user/containerd.service
...
[INFO] Installed containerd.service successfully.
[INFO] To control containerd.service, run: `systemctl --user (start|stop|restart) containerd.service`
[INFO] To run containerd.service on system startup, run: `sudo loginctl enable-linger testuser`

[INFO] Use `nerdctl` to connect to the rootless containerd.
[INFO] You do NOT need to specify $CONTAINERD_ADDRESS explicitly.

The usage of containerd-rootless-setuptool.sh is almost same as dockerd-rootless-setuptool.sh .

Resource limitation flags such as nerdctl run --memory require systemd and cgroup v2: https://rootlesscontaine.rs/getting-started/common/cgroup2/

Client (nerdctl)

Just execute nerdctl. No need to specify the socket address manually.

$ nerdctl run -it --rm alpine

Depending on your kernel version, you may need to enable FUSE-OverlayFS or set export CONTAINERD_SNAPSHOTTER=native. (See below.)

Add-ons

BuildKit

To enable BuildKit, run the following command:

$ containerd-rootless-setuptool.sh install-buildkit

Snapshotters

OverlayFS

The default overlayfs snapshotter only works on the following hosts:

  • Any distro, with kernel >= 5.13
  • Non-SELinux distro, with kernel >= 5.11
  • Ubuntu since 2015

For other hosts, fuse-overlayfs snapshotter needs to be used instead.

FUSE-OverlayFS

To enable fuse-overlayfs snapshotter, run the following command:

$ containerd-rootless-setuptool.sh install-fuse-overlayfs

Then, add the following config to ~/.config/containerd/config.toml, and run systemctl --user restart containerd.service:

[proxy_plugins]
  [proxy_plugins."fuse-overlayfs"]
      type = "snapshot"
# NOTE: replace "1000" with your actual UID
      address = "/run/user/1000/containerd-fuse-overlayfs.sock"

The snapshotter can be specified as $CONTAINERD_SNAPSHOTTER.

$ export CONTAINERD_SNAPSHOTTER=fuse-overlayfs
$ nerdctl run -it --rm alpine

If fuse-overlayfs does not work, try export CONTAINERD_SNAPSHOTTER=native.

Stargz Snapshotter

Stargz Snapshotter enables lazy-pulling of images.

To enable Stargz snapshotter, run the following command:

$ containerd-rootless-setuptool.sh install-stargz

Then, add the following config to ~/.config/containerd/config.toml and run systemctl --user restart containerd.service:

[proxy_plugins]
  [proxy_plugins."stargz"]
      type = "snapshot"
# NOTE: replace "1000" with your actual UID
      address = "/run/user/1000/containerd-stargz-grpc/containerd-stargz-grpc.sock"

The snapshotter can be specified as $CONTAINERD_SNAPSHOTTER.

$ export CONTAINERD_SNAPSHOTTER=stargz
$ nerdctl run -it --rm ghcr.io/stargz-containers/alpine:3.10.2-esgz

See https://github.com/containerd/stargz-snapshotter/blob/main/docs/pre-converted-images.md for the image list.

bypass4netns

⚡ Requirement nerdctl >= 0.17

bypass4netns(https://github.com/rootless-containers/bypass4netns) is an accelerator for rootless networking.

This improves outgoing or incoming (with --publish option) networking performance.

The performance benchmark with iperf3 on Ubuntu 21.10 on Hyper-V VM is shown below.

iperf3 benchmark without bypass4netns with bypass4netns
container -> host 0.398 Gbps 42.2 Gbps
host -> container 20.6 Gbps 47.4 Gbps

This benchmark can be reproduced with https://github.com/rootless-containers/bypass4netns/blob/f009d96139e9e38ce69a2ea8a9a746349bad273c/Vagrantfile

Acceleration with bypass4netns is available with --label nerdctl/bypass4netns=true. You also need to have bypass4netnsd (bypass4netns daemon) to be running. Example

$ containerd-rootless-setuptool.sh install-bypass4netnsd
$ nerdctl run -it --rm -p 8080:80 --label nerdctl/bypass4netns=true alpine

More detail is available at https://github.com/rootless-containers/bypass4netns/blob/master/README.md

Configuring RootlessKit

Rootless containerd recognizes the following environment variables to configure the behavior of RootlessKit:

  • CONTAINERD_ROOTLESS_ROOTLESSKIT_STATE_DIR=DIR: the rootlesskit state dir. Defaults to $XDG_RUNTIME_DIR/containerd-rootless.
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_NET=(slirp4netns|vpnkit|lxc-user-nic): the rootlesskit network driver. Defaults to "slirp4netns" if slirp4netns (>= v0.4.0) is installed. Otherwise defaults to "vpnkit".
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_MTU=NUM: the MTU value for the rootlesskit network driver. Defaults to 65520 for slirp4netns, 1500 for other drivers.
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_PORT_DRIVER=(builtin|slirp4netns): the rootlesskit port driver. Defaults to "builtin".
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_SLIRP4NETNS_SANDBOX=(auto|true|false): whether to protect slirp4netns with a dedicated mount namespace. Defaults to "auto".
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_SLIRP4NETNS_SECCOMP=(auto|true|false): whether to protect slirp4netns with seccomp. Defaults to "auto".
  • CONTAINERD_ROOTLESS_ROOTLESSKIT_DETACH_NETNS=(auto|true|false): whether to launch rootlesskit with the "detach-netns" mode. Defaults to "auto", which is resolved to "true" if RootlessKit >= 2.0 is installed. The "detached-netns" mode accelerates nerdctl (pull|push|build) and enables nerdctl run --net=host, however, there is a relatively minor drawback with BuildKit prior to v0.13: the host loopback IP address (127.0.0.1) and abstract sockets are exposed to Dockerfile's "RUN" instructions during nerdctl build (not nerdctl run). The drawback is fixed in BuildKit v0.13. Upgrading from a prior version of BuildKit needs removing the old systemd unit: containerd-rootless-setuptool.sh uninstall-buildkit && rm -f ~/.config/buildkit/buildkitd.toml

To set these variables, create ~/.config/systemd/user/containerd.service.d/override.conf as follows:

[Service]
Environment=CONTAINERD_ROOTLESS_ROOTLESSKIT_DETACH_NETNS="false"

And then run the following commands:

systemctl --user daemon-reload
systemctl --user restart containerd

Troubleshooting

Hint to Fedora users

Rootlesskit Network Design

In detach-netns mode:

  • Network namespace is detached and stored in $ROOTLESSKIT_STATE_DIR/netns.
  • The child command executes within the host's network namespace, allowing actions like pull and push to happen in the host network namespace.
  • For creating and configuring the container's network namespace, the child command switches temporarily to the relevant namespace located in $ROOTLESSKIT_STATE_DIR/netns. This ensures necessary network setup while maintaining isolation in the host namespace.

rootlessKit-network-design.png

  • Rootlesskit Parent NetNS and Child NetNS are already configured by the startup script containerd-rootless.sh
  • Rootlesskit Parent NetNS is the host network namespace
  • step1: nerdctl calls containerd in the host network namespace.
  • step2: containerd calls runc in the host network namespace.
  • step3: runc creates container with dedicated namespaces (e.g network ns) in the Parent netns.
  • step4: runc nsenter Rootlesskit Child NetNS before triggering nerdctl ocihook.
  • step5: nerdctl ocihook module leverages CNI.
  • step6: CNI configures container network namespace: create network interfaces eth0 -> veth0 -> nerdctl0.