deb: fix failure on purge (td-agent v4)

[kenhys]

According to fluent/fluentd-website#266, purge failure is reported.

vagrant@vagrant:~$ apt list --installed | grep fluen

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

fluent-package/now 5.0.0-1 amd64 [installed,local]
fluentd-apt-source/focal,now 2020.8.25-1 all [installed]
vagrant@vagrant:~$ apt list --installed | grep td-agent

WARNING: apt does not have a stable CLI interface. Use with caution in scripts.

vagrant@vagrant:~$ sudo apt purge td-agent
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following packages will be REMOVED:
  td-agent*
0 upgraded, 0 newly installed, 1 to remove and 194 not upgraded.
After this operation, 0 B of additional disk space will be used.
Do you want to continue? [Y/n] 
(Reading database ... 50480 files and directories currently installed.)
Purging configuration files for td-agent (4.5.0-1) ...
td-agent:x:113:117::/var/lib/td-agent:/usr/sbin/nologin
userdel: user td-agent is currently used by process 21510
dpkg: error processing package td-agent (--purge):
 installed td-agent package post-removal script subprocess returned error exit status 8
Errors were encountered while processing:
 td-agent
E: Sub-process /usr/bin/dpkg returned an error code (1)

[kenhys]

Verified with the following situation:

 dpkg -l|grep -E "td-agent|fluent"
ii  fluent-package                        5.0.0-1                                 amd64        Treasure Agent: A data collector for Treasure
 Data
rc  td-agent                              4.5.0-1                                 amd64        Treasure Agent: A data collector for Treasure
 Data

Then execute sudo apt purge td-agent.

td-agent v4.5.0 is assumed that td-agent service must be stopped before purging td-agent package.

. /usr/share/debconf/confmodule

if [ "$1" = "purge" ]; then
        rm -f /etc/default/td-agent
        dpkg-statoverride --list /etc/td-agent > /dev/null && \
                dpkg-statoverride --remove /etc/td-agent
        rm -f /etc/td-agent/td-agent.conf
        rm -rf /etc/td-agent
        dpkg-statoverride --list /var/run/td-agent > /dev/null && \
                dpkg-statoverride --remove /var/run/td-agent
        rm -f /var/run/td-agent/*
        rm -rf /var/run/td-agent
        dpkg-statoverride --list /var/log/td-agent > /dev/null && \
                dpkg-statoverride --remove /var/log/td-agent
        rm -rf /var/log/td-agent/buffer
        rm -rf /var/log/td-agent/*
        rm -rf /var/log/td-agent

        getent passwd td-agent && userdel -r td-agent
fi

[kenhys]

On jammy, td-agent is marked as rc. on the contract to it, td-agent is marked as ii on focal.

jammy:

$ dpkg -l |grep td-agent
rc  td-agent                              4.5.0-1                                 amd64        Treasure Agent: A data collector for Treasure
 Data
vagrant@ubuntu2204:~$ sudo apt purge td-agent
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following packages will be REMOVED:
  td-agent*
0 upgraded, 0 newly installed, 1 to remove and 0 not upgraded.
After this operation, 0 B of additional disk space will be used.
Do you want to continue? [Y/n] y
(Reading database ... 120911 files and directories currently installed.)
Purging configuration files for td-agent (4.5.0-1) ...
td-agent:x:114:121::/var/lib/td-agent:/usr/sbin/nologin
userdel: user td-agent is currently used by process 39797
dpkg: error processing package td-agent (--purge):
 installed td-agent package post-removal script subprocess returned error exit status 8
Errors were encountered while processing:
 td-agent
E: Sub-process /usr/bin/dpkg returned an error code (1)

focal:

$ sudo apt purge td-agent
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following packages will be REMOVED:
  td-agent*
0 upgraded, 0 newly installed, 1 to remove and 1 not upgraded.
After this operation, 15.4 kB disk space will be freed.
Do you want to continue? [Y/n] y
(Reading database ... 121430 files and directories currently installed.)
Removing td-agent (5.0.0-1) ...
(Reading database ... 121427 files and directories currently installed.)
Purging configuration files for td-agent (5.0.0-1) ...

[kenhys]

If you purge td-agent 5.0.0, no problem.

$ sudo apt purge td-agent
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following packages will be REMOVED:
  td-agent*
0 upgraded, 0 newly installed, 1 to remove and 0 not upgraded.
After this operation, 15.4 kB disk space will be freed.
Do you want to continue? [Y/n] y
(Reading database ... 120972 files and directories currently installed.)
Removing td-agent (5.0.0-1) ...
(Reading database ... 120969 files and directories currently installed.)
Purging configuration files for td-agent (5.0.0-1) ...

[daipom]

Hmm, can we really purge v4 safely?

If a user upgrades to v5 from v4, I suppose that the v4 package should not be purged.

The focal environment where the package is upgraded to v5 from v4:

$ dpkg -l td-agent
Desired=Unknown/Install/Remove/Purge/Hold
| Status=Not/Inst/Conf-files/Unpacked/halF-conf/Half-inst/trig-aWait/Trig-pend
|/ Err?=(none)/Reinst-required (Status,Err: uppercase=bad)
||/ Name           Version      Architecture Description
+++-==============-============-============-==================================================
rc  td-agent       4.5.0-1      amd64        Treasure Agent: A data collector for Treasure Data

$ dpkg -L td-agent
/etc
/etc/logrotate.d
/etc/logrotate.d/td-agent
/etc/td-agent
/etc/td-agent/td-agent.conf
/opt

$ ls -la /etc/td-agent
lrwxrwxrwx 1 root root 11 Aug  8 08:24 /etc/td-agent -> /etc/fluent

$ ls -la /etc/fluent/
drwxr-xr-x  3 td-agent td-agent 4096 Aug  8 08:24 .
drwxr-xr-x 99 root     root     4096 Aug  8 08:24 ..
lrwxrwxrwx  1 root     root       25 Aug  8 08:24 fluentd.conf -> /etc/fluent/td-agent.conf
drwxr-xr-x  2 root     root     4096 Aug  8 08:24 plugin
-rw-r--r--  1 root     root     2382 Aug  8 08:24 td-agent.conf

For example, I suppose that we should not purge the config file: /etc/td-agent/td-agent.conf because it is still used by v5.

In addition, I wonder if this v4 postrm script would break the migrated environment...

fluent-package-builder/td-agent/templates/package-scripts/td-agent/deb/postrm

Lines 7 to 24 in ac9ca53

	if [ "$1" = "purge" ]; then
	rm -f /etc/default/<%= project_name %>
	dpkg-statoverride --list /etc/<%= project_name %> > /dev/null && \
	dpkg-statoverride --remove /etc/<%= project_name %>
	rm -f /etc/<%= project_name %>/<%= project_name %>.conf
	rm -rf /etc/<%= project_name %>
	dpkg-statoverride --list /var/run/<%= project_name %> > /dev/null && \
	dpkg-statoverride --remove /var/run/<%= project_name %>
	rm -f /var/run/<%= project_name %>/*
	rm -rf /var/run/<%= project_name %>
	dpkg-statoverride --list /var/log/<%= project_name %> > /dev/null && \
	dpkg-statoverride --remove /var/log/<%= project_name %>
	rm -rf /var/log/<%= project_name %>/buffer
	rm -rf /var/log/<%= project_name %>/*
	rm -rf /var/log/<%= project_name %>
	getent passwd <%= project_name %> && userdel -r <%= project_name %>
	fi

[daipom]

I confirmed we can purge td-agent v4 without the error if we stop the service in advance.
However, as I am concerned, files are deleted unexpectedly, corrupting the environment.

The focal environment where the package is upgraded to v5 from v4:

$ sudo systemctl stop fluentd
$ sudo apt purge td-agent
Reading package lists... Done
Building dependency tree       
Reading state information... Done
The following packages will be REMOVED:
  td-agent*
0 upgraded, 0 newly installed, 1 to remove and 194 not upgraded.
After this operation, 0 B of additional disk space will be used.
Do you want to continue? [Y/n] 
(Reading database ... 50480 files and directories currently installed.)
Purging configuration files for td-agent (4.5.0-1) ...
td-agent:x:113:117::/var/lib/td-agent:/usr/sbin/nologin
userdel: group td-agent is the primary group of another user and is not removed.
userdel: td-agent mail spool (/var/mail/td-agent) not found

# It seems to be no problem that the user `td-agent` is removed.
vagrant@vagrant:~$ getent passwd td-agent
(empty)

vagrant@vagrant:~$ getent passwd _fluentd
_fluentd:x:113:117::/var/lib/fluent:/usr/sbin/nologin

# The config file and the symbolic link are removed.
vagrant@vagrant:~$ la /etc/td-agent
ls: cannot access '/etc/td-agent': No such file or directory

vagrant@vagrant:~$ ls -la /etc/fluent
drwxr-xr-x  3 _fluentd td-agent 4096 Aug  8 09:09 .
drwxr-xr-x 99 root     root     4096 Aug  8 09:09 ..
lrwxrwxrwx  1 root     root       25 Aug  8 09:07 fluentd.conf -> /etc/fluent/td-agent.conf
drwxr-xr-x  2 root     root     4096 Aug  8 09:07 plugin

# The log/buffer files and the symbolic link are cleared.
vagrant@vagrant:~$ ls -la /var/log/td-agent
ls: cannot access '/var/log/td-agent': No such file or directory

vagrant@vagrant:~$ ls -la /var/log/fluent/
drwxr-xr-x 2 _fluentd td-agent 4096 Aug  8 09:09 .
drwxrwxr-x 8 root     syslog   4096 Aug  8 09:09 ..

[daipom]

This is specification of Fluentd DEB package.
Some old files should remain after update, so the old package remains as rc, that means it is uninstalled, but some files are remained.
We should not purge the old package.