Bug#881329: systemd service
Hi again,
I had to add
RestartPreventExitStatus=1
in [Service], because on certain failures (e.g. broken permissions in
the lock directory) mailmanctl forks and exits (which systemd considers
a successful start), and then the fork exits with status code 1.
Because it was started "successfully", systemd will trigger a restart
without limit...
It would be better if mailmanctl supported a flag to not fork mailmanctl
itself, but just run in the foreground (no need to close std{in,out,err}
either). systemd (and other supervising init systems) can handle that
way better.
cheers,
Stefan
Bug#881329: systemd service
Package: mailman Version: 1:2.1.23-1+deb9u1 Hi, it would be nice if the mailman package offered a real systemd service; that way it could be automatically restarted on crashes (e.g. https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=505638), and `systemctl list-units --failed` would complain properly when it isn't working. For this I suggest these two files: /usr/lib/tmpfiles.d/mailman.conf --- # path mode uid gid d /run/mailman 0755 list list d /run/lock/mailman 2775 root list --- (dh_installinit should create these automatically in .postinst when systemd is active) /lib/systemd/system/mailman.service --- [Unit] Description=Mailman Master Queue Runner After=network.target [Service] Type=forking PIDFile=/var/run/mailman/mailman.pid ExecStart=/usr/lib/mailman/bin/mailmanctl -s start ExecStop=/usr/lib/mailman/bin/mailmanctl stop ExecReload=/usr/lib/mailman/bin/mailmanctl restart Restart=always RestartSec=3s [Install] WantedBy=multi-user.target Alias=mailman-qrunner.service --- mailmanctl checks for the existence of the site list on start itself, I don't see any need to keep/clone this logic. In [Unit] one could add: --- ConditionFileIsExecutable=/usr/bin/python --- to replicate the check from the init script. cheers, Stefan
