Hello,
For what it's worth, still experiencing these symptoms with 2.2.5.
Thanks, Simon.
On Thu, 2013-08-01 at 15:02 +0100, Simon Fraser wrote:
On Thu, 2013-08-01 at 15:09 +0300, Timo Sirainen wrote:
On Fri, 2013-07-26 at 11:28 +0100, Simon Fraser wrote:
I am running dovecot 2.2.2 with tcp based replication, and experiencing some duplicated emails.
doveconf -n
output is below.Are both of the servers using the same mailbox format?
Yes, they are.
doveconf -n | grep mail_location
on both nodes gives: mail_location = maildir:~/mail:INBOX=~/mail/INBOXI've also tried with mdbox.
Connect with a mail client, and delete the message - without delayed expunge. So, for example, mutt (press 'd' then '$' to sync the mailbox), or Evolution set to immediately delete.
Can you reproduce this by disabling automatic replication (e.g. just remove replicator from mail_plugins) and running dsync manually? So basically delete the message, then run:
doveadm sync -r rawlog -u user@domain -d
If that reproduces it, send the rawlog to me.
I can't reproduce it with that, sorry. Message deletion appears to work that way, and the command only produces I/O leak errors:
dsync-local(srf@sanger.ac.uk): Warning: I/O leak: 0x7f00e50cc960 (line 341, fd 9)
Using mail delivery and imap connections to node 'a' only, I have tried:
- mail delivered, connecting/deleting/expunging, then sync
- mail delivered, syncing, connecting/deleting/expunging, not syncing, then checking
- mail delivered, syncing, connecting/deleting/expunging, syncing
None cause the message to reappear. I tried initiating the sync from both nodes, just in case.
Leaving the replication plugin enabled causes it to happen, though, so perhaps my timing is off.
Other things I have discovered by running 'watch' or a while loop, both grepping the mail_location for my test subject line:
A message is visible over imap before being written to mail_location. If I can manage to delete it in this time (there isn't much time!) then the message does not reappear.
When the message reappears it gets a different filename (please excuse the grep output on the end): cur/1375364237.M969208P14576.intmail3a,S=2672,W=2731:2,:Subject: dup test
is replaced with: cur/1375364242.M782761P838.intmail3b,S=2672,W=2731:2,:Subject: dup test
This occurs less than a second after the expunge. There's only a log entry on intmail3a, not 3b:
Aug 1 14:37:26 intmail3a dovecot: imap(srf): Warning: /mail/spool/s/srf/mail/INBOX/dovecot-uidlist: Duplicate file entry at line 105: 1375364237.M969208P14576.intmail3a,S=2672,W=2731 (uid 772051 -> 772052)
I'm happy to run as many diagnostics as required, just let me know what you'd like.
Simon.
-- The Wellcome Trust Sanger Institute is operated by Genome Research Limited, a charity registered in England with number 1021457 and a company registered in England with number 2742969, whose registered office is 215 Euston Road, London, NW1 2BE.