[Dovecot] [dovecot] Dovecot Migration so far

[Mark Moseley]

On Wed, Jul 7, 2010 at 12:22 PM, Timo Sirainen <tss at iki.fi> wrote:
> On Wed, 2010-07-07 at 10:19 -0700, Mark Moseley wrote:
>> On Wed, Jul 7, 2010 at 8:08 AM, Timo Sirainen <tss at iki.fi> wrote:
>> > On Fri, 2010-04-02 at 11:35 -0700, Mark Moseley wrote:
>> >> Apr 02 13:25:24 imap(mail at box): Error:
>> >> creat(/var/tmp/Indexes/14/75/mail at box/dovecot.mailbox.log) failed: No
>> >> such file or directory
>> >
>> > Finally fixed: http://hg.dovecot.org/dovecot-2.0/rev/6571b3a9c44f
>>
>> Is that revision pre-rc1 or contained in it?
>
> It's post-rc1. I fixed it a few minutes before sending the mail.

Looks good. I haven't seen any yet.


>> I've been running beta5 for a while and not been seeing those. The
>> closest thing I get now is "Error: FETCH for mailbox INBOX UID 4322
>> failed to read message input: Stale NFS file handle", but that's
>> obviously not Dovecot's fault.
>
> You should be running the new director I think. :)

I've only got one instance of this error so far and none of the 'got
too little data' errors. I've been overwhelmingly happy with just
running things on NFS so far. Usually my mail users (and there's
millions of them) go berserk over mail issues but there's been an
almost disconcerting quiet regarding IMAP ever since we moved to
Dovecot.


The only oddities I've found in my logs so far since upgrading to rc2
(or at least I didn't notice before rc2) are these. These are all just
very minor (between them, maybe 10-15 mentions in 250k lines of logs).

1. Jul 20 12:17:30 imap-login: Info: Internal login failure (auth
failed, 1 attempts): user=<user at mailbox.com>, method=PLAIN,
rip=x.x.x.x, lip=192.168.152.37, mpid=0, TLS

2. Jul 20 12:17:41 auth: Error: Master requested auth for nonexisting
client 16380

3. Just a single instance of: Jul 20 09:40:38
imap(username at mailbox.com): Warning: /path/to/Maildir/dovecot-uidlist:
Duplicate file entry at line 23:
1279633198.H285683P32723.servername,S=13870:2,ad (uid 669 -> 670)


In both cases, there's only a few of each, in the space of 250k lines
of logs, so very rare, but slightly puzzling. I'm guessing in #2, the
process must've died/gotten killed. Looking at the source for #1, I
have no clue (nor do we use SASL -- everything is using mysql for
auth). #3 might be chalked up to just using NFS?