Dovecot dotlock error - locking set to fcntl ??
 

We are currently running Dovecot 1.2 on a Centos 5.5 machine for a production workload of around 56,000 - 60,000 pop/imap logins per hour of the day (totaling a million something a day). The server has it's storage hosted on nfs. This server is usually humming along just fine until we get the following error in the mail syslog:

Created dotlock file's timestamp is different than current time (1281964699 vs 1281964649): /mail/user/Mai
ldir/dovecot-uidlist

We see a couple of these errors and the server explodes with a load average of over 300, which drops back down immediately but definitely causes a small outage. The strangest part is that we are NOT using dotlock as a locking mechanism on this server. We are also not using dotlock as a locking mechanism for any of the servers that access this nfs storage, rather we use fcntl. There was a time when we tested dotlock on the servers in our mail cluster, but we are not sure why these errors are appearing now after running without dotlock for so long.

Thanks for your help!

Sounds like something else is accessing this file other than just dovecot, Since your running NFS I'm going to assume you have other Incoming SMTP servers (spamfilters too) that might be changing the users mail spool. I would recommend going on all servers that access these mail spools and using the lsof command to determine what else is accessing the spool.

It could be as simple as someone using pine.