On Wed, 2008-10-01 at 21:15 -0400, Karsten Künne wrote:
we recently upgraded dovecot from 1.0.5 to 1.1.3+latest patches. After the upgrade I saw some errors in the logs and I'm not sure whether they're a cause for concern. Our setup is 2 servers running openSUSE 11.0 and a NetApp Filer as storage backend. Following is the config:
So it's possible that two users are accessing the same mailbox from two different servers at the same time?
It was working the same way with v1.0 too? Without errors? That sounds strange, since that kind of a setup should have been much more broken with v1.0.
mail_location: maildir:~/.Maildir:CONTROL=/var/mailctl/%n:INDEX=/var/mailctl/%n
I guess /var/mailctl is also on netapp?
mail_nfs_storage: yes mail_nfs_index: yes
These should have prevented these errors, but I guess they don't fully work everywhere then..
Oct 1 07:39:27 host1 dovecot: IMAP(user1): Invalid transaction log size (83856 vs 83160): /var/mailctl/corey/.INBOX/dovecot.index.log ..
None of these should ever happen.
Oct 1 09:14:55 host1 dovecot: IMAP(user2): Broken file /var/mailctl/user2/.INBOX/dovecot-uidlist line 15299: UIDs not ordered (206464 > 206464)
Oct 1 09:16:43 host1 dovecot: IMAP(user3): Broken file /var/mailctl/user3/.Trash/dovecot-uidlist line 22: UID larger than next_uid (153866 >= 153866)
These especially could cause visible problems to clients.
Oct 1 07:29:05 host1 dovecot: IMAP(user4): nfs_flush_fcntl: fcntl(/var/mailctl/user4/.INBOX/dovecot.index.cache, F_RDLCK) failed: Interrupted system call
Do you get a lot of these? This means it was trying to flush NFS read cache by read-locking the file, but it failed after waiting for 60 seconds. All of these locks should be really short, so if it's timing out things aren't working very well.