Thanks for responding - it's nice to know there isn't a plot to give me the silent treatment!
I had already looked at the email logs - that was how I figured out that time had done a weird thing and that was the reason dovecot had shut down. As for the two files you mentioned, there are no files named like that anywhere on my server. The only file starting with "dovecot" and sounding like a log of any kind is /home/user/mail/.imap/INBOX/dovecot.index.log. (The "INBOX" in that path is not actually used for any mail, because I don't use IMAP). The timestamp on that file is the night before the incident happened (matching one of the last times I checked mail successfully, although interestingly not the very last time), so that's not relevant.
The maillog entries that are relevant are:
Code:
Jun 6 01:44:21 myVPS postfix/postfix-script: stopping the Postfix mail system
Jun 6 01:44:21 myVPS postfix/master[1607]: terminating on signal 15
Jun 6 01:44:21 myVPS dovecot: Killed with signal 15
Jun 6 14:06:35 myVPS dovecot: Dovecot v1.0.7 starting up
Jun 6 14:07:03 myVPS postfix/postfix-script: starting the Postfix mail system
Jun 6 14:07:06 myVPS postfix/master[722]: daemon started -- version 2.3.3, configuration /etc/postfix
Jun 6 05:07:00 myVPS dovecot: Time just moved backwards by 32417 seconds. This might cause a lot of problems, so I'll just kill myself now. http://wiki.dovecot.org/TimeMovedBackwards
When I discovered it a couple hours later, I restarted dovecot, but after that I was always refused authentication, with a maillog entry like this:
Code:
Jul 1 23:04:14 myVPS dovecot: pop3-login: Disconnected: user=<myuser>, method=PLAIN, rip=::ffff:IP.of.my.ISP, lip=::ffff:IP.of.my.server, TLS
I have no idea why postifx and dovecot shut down at 1:44, or why sometime while they were shut down the system time jumped forward (presumably by the same 9 hours that it then jumped backward right after postfix and dovecot restarted themselves). The server did not reboot.