On Dec 3, 2006, at 3:06 PM, Timo Sirainen wrote:
On Sun, 2006-12-03 at 14:53 -0500, Sean Kamath wrote:
- Disconnected for inactivity: This has sometimes been a bug in Dovecot, but I'm not aware of any such bugs currently. It could also just be a network problem.
I recently got a set of duplicate emails (including this one). I did have some connectivity problems late last night, and this I saw in the log file:
And you delete mails from server after downloading?
I'm not. I mean, that's sort of the point of imap, right? :-) :-)
Dec 1 10:42:04 stove dovecot: [ID 107833 mail.info] IMAP(kamath): Disconnected for inactivity
So, possibly network problem..
Most definately. With UW Imap (used prior to dovecot), having mail
on the slow NFS link (10 Meg link) would occasionally timeout on
logins and you'd also get duplicate email. Moving to local storage
*mostly* fixed that. Dovecot has been MUCH better, but it's still
happening.
Still and all, I'm curious how any sort of network problem could
cause this. With UW, we'd move mail out of the spool file to the
mbox in the home dir, and if the network connection died (because it
took too long to copy the spool file), you'd end up with mail in the
spool file and the mbox. But now we're using the spool file directly.
The next time I notice it, I'll confirm (again) that it's actually
duplicated *on the server* and *in the mail spool file* (we use mbox).
Dec 1 10:42:07 stove dovecot: [ID 107833 mail.error] IMAP (kamath): X- IMAPbase uid-last unexpectedly lost in mbox file /var/mail/kamath
But why did this happen? Is this 1.0.rc15? Is something else modifying the mbox files also?
Aside from sendmail and the LDA, no, nothing should be modifying the
mbox file.
If the problem isn't disconnection, then it could be that the UIDs changed, and that could be related to the above error.
Gack. rc7. I guess I haven't upgraded in a while. Frankly, rc7 has
been working so well, and what with the recent spate of activity,
I've been waiting for things to settle down a bit -- but I didn't
think I was that far back.
Sean