[Dovecot] Dovecot 1.2.9 Crash, NFS
lukas.mueller at newmedia.ch
Mon Mar 26 18:25:54 EEST 2012
Thanks for the quick answer.
I realised, that the error didn't occur since quite a while, opposed to what our client suggested.
Back then I activated the two workarounds (imap_client_workarounds = outlook-idle delay-newmail) and increased mail_max_userip_connections for IMAP.
Is it possible that those could have improved the situation?
For now I don't have a way of reproducing the problem, so I will have to wait for an error to happen.
Until then I will consider the problem as "Solved until happens again ;-) ", since the last error occured a while back (as mentioned above).
>> Mar 6 08:26:31 mail02 dovecot: IMAP(user at example.com): fdatasync(/data/vmail/example.com/user/dovecot-uidlist) failed: Input/output error
>> Mar 6 08:42:29 mail02 dovecot: IMAP(user at example.com): Maildir /data/vmail/example.com/user: Expunged message reappeared, giving a new UID (old uid=1522, file=1326961561.V15I4d8562M567017.mail02:2,Sad)
>> Mar 6 08:42:29 mail02 dovecot: IMAP(user at example.com): Maildir /data/vmail/example.com/user: Expunged message reappeared, giving a new UID (old uid=1523, file=1326705103.V15I90105M613353.mail01:2,Sad)
>> Mar 6 08:42:29 mail02 dovecot: IMAP(user at example.com): /data/vmail/example.com/user/dovecot-uidlist: Duplicate file entry at line 4: 1326961561.V15I4d8562M567017.mail02:2,Sad (uid 1522 -> 1598)
>> My suspicion/speculation what happens is the following:
>> Multiple users are accessing the Mailbox from their offices (all on the same server), one (or more) uses the Webmail or accesses the Mailbox from a different IP.
>> Somehow this leads to problems with Locks on NFS, which leads to the crash.
>Yes, most likely this is what's happening. Although your errors are more severe than what normally happens. I guess your NFS server is also partially to blame (microsecond resolution timestamps are at least helpful).
I had a quick look a tour NFS (NetApp), but didn't find anything useful.
In case the problem persists, I will check with the coworker responsible for NetApp.
I will check what sort of locking is used by postfix, since I'm not sure if postfix and dovecot are configured to use the same mechanisms.
If not, I think it could be part of the Problem.
>> I have no idea how to solve this problem and any help is greatly appreciated.
>The only way to fully fix this is: http://wiki2.dovecot.org/Director
Unfortunately that is not an option right now, but I will keep it in mind.
More information about the dovecot