FC16 - spamassassin fails on boot, succeeds when restarted

[modal1]

Not having much luck finding a solution for this issue anywhere. Perhaps someone here can shed some light.

After rebooting:

/var/log/messages:

sendmail[1535]: WARNING: Xspamassassin: local socket name /var/run/spamass-milter/spamass-milter.sock missing

but the file exists:

srwxr-xr-x 1 sa-milt sa-milt 0 Aug 12 14:47 /var/run/spamass-milter/spamass-milter.sock

/var/log/maillog:

sendmail[1934]: q7CLPViK001934: Milter (spamassassin): local socket name /var/run/spamass-milter/spamass-milter.sock unsafe
sendmail[1934]: q7CLPViK001934: Milter (spamassassin): to error state

If I restart spamassassin and spamass-milter manually, everything works fine - no errors.


Thanks.

[William Haller]

/var/run is a tmpfs. It is possible that it is not yet created when spamassassin tries to create the socket and exists by the time you do a restart. Thus the first run through of sendmail fails. You might also need to add a dependency of SA in the sendmail run tree so it for sure starts first.

The systemd modification to start which still doesn't appear to be complete has several ordering issues. This may be one of them.

[modal1]

I think you've done it!

Adding

After = spamassassin.service spamass-milter.service

to /lib/systemd/system/sendmail.service

still generated the same errors in /var/log/messages after a reboot but spamassassin works without having to be restarted!

I am still seeing the following in the messages log:

systemd[1]: PID file /var/run/akmodsd.pid not readable (yet?) after start.
systemd[1]: Unit akmods.service entered failed state.
systemd[1]: PID file /run/nmbd.pid not readable (yet?) after start.
systemd[1]: PID file /run/smbd.pid not readable (yet?) after start.
systemd[1]: PID file /var/run/httpd/httpd.pid not readable (yet?) after start.

despite the fact that all these processes (except akmods) run without issue.

I can live with the extra log entries and the akmods failure I can deal with later.

Thank you!