Sergey wrote:
Timo Sirainen wrote:
On 25.7.2007, at 17.18, Sergey wrote:
dovecot: Jul 25 01:05:19 Error: IMAP(seriv): Corrupted index cache file /var/spool/imap/seriv/.imap/.community-lists-altlinux-org/dovecot.index.cache:
Corrupted cach ed message_part data (physical_pos less than expected) Hmm. This is probably a new bug..
deliver(seriv): Jul 25 04:17:26 Panic: seq = 1587, rec->uid = 0, first_new_seq = 1587, records = 1586 But this crash is in mbox rewriting code which hasn't really changed since v1.0.
Can you easily reproduce both of these? How? Are you using deliver also with v1.0? Doesn't it crash?
I have there 2 dovecots and 2 dovecot-sieve plugins installed in parallel. So, it may be just error in installing, so that some file confusion. My previous attempt, when I forgot to change in dovecot11.conf line '/usr/lib64/dovecot11/modules/lda' and it was pointing to lda of dovecot-1.0.2, has a very reproducible problem: deliver crashed at every message :-) Dovecot-1.0.2 has the same sieve script and works perfect. I had during last hours only 5 backtraces, with corresponding lines in postfix' log like the following:
Jul 25 08:11:12 mx01 postfix/qmgr[12199]: 5B8CF534C2E: from=sisyphus-bounces@lists.altlinux.org, size=5343, nrcpt=1 (queue active) Jul 25 08:11:12 mx01 postfix/pipe[14290]: 5B8CF534C2E: to=seriv@parkheights.dyndns.org, relay=dovecot11, delay=0.32, delays=0.29/0/0/0.03, dsn=5.3.0, status=bounced (Command died with signal 6: "/usr/lib64/dovecot11/deliver") Jul 25 08:11:12 mx01 postfix/cleanup[14198]: 9D88D534BD9: message-id=20070725121112.9D88D534BD9@mx01.menlo Jul 25 08:11:12 mx01 postfix/qmgr[12199]: 9D88D534BD9: from=<>, size=7080, nrcpt=1 (queue active) Jul 25 08:11:12 mx01 postfix/bounce[15778]: 5B8CF534C2E: sender non-delivery notification: 9D88D534BD9 Jul 25 08:11:12 mx01 postfix/qmgr[12199]: 5B8CF534C2E: removed
Total number of messages delivered by dovecot11-deliver are 240, so I'm not sure I can reproduce it quickly. I'll try to get coredumps of deliver. I have 4 new backtraces logged into /tmp/dovecot.log. I have added line "mail_drop_priv_before_exec: yes" to dovecot11.conf, and verified that userdb, both static and from passwd file, returns user homes, and these homes are owned by user vmail specified in /etc/postfix/master.cf to run dovecot11 service. And can't find any core files written anywhere. What I'm missing? How to get coredumps of delivery agent?
I have read once more wiki pages about using dovecot delivery, and discovered that postfix configuration is missing a line
dovecot11_destination_recipient_limit = 1
May be, it's related to the problem? It looks like last cases of the problem occur at the time of bursts of mails arriving to postfix. I have added this line and will continue testing.
-- Sergey.