im not sure if i enabled core dumps

im using Centos 5.5 i echoed "'DAEMON_COREFILE_LIMIT="unlimited"' >> /etc/sysconfig/dovecot" and restarted dovecot no core files

but i dont have any killed messages in my logs anyway like: dovecot: Apr 23 11:16:05 Error: child 86116 (imap) killed with signal 11 only " (Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda")"

2010/10/4 spamvoll@googlemail.com:

hard to reproduce.. happens only with some mail.

with mail_debug = yes i cant see any errors, but the mail get delivered 2 times ?

Oct  4 20:17:39 imap dovecot: lda(woelky@example.com): sieve: msgid=374f9ab6-5153-47b3-bbfc-cde0705c55d1@email.android.com: stored mail into mailbox 'INBOX' Oct  4 20:17:39 imap postfix/pipe[14085]: 411E6F5572: to=woelky@example.com, relay=dovecot, delay=1, delays=0.53/0.03/0/0.44, dsn=2.0.0, status=sent (delivered via dovecot service) Oct  4 20:17:40 imap postfix/pipe[14082]: 411E6F5572: to=malte.woelky@example.com, relay=dovecot, delay=1.1, delays=0.53/0.02/0/0.53, dsn=5.3.0, status=bounced (Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda")

2010/10/4 Timo Sirainen tss@iki.fi:

...

On Mon, 2010-10-04 at 14:36 +0200, spamvoll@googlemail.com wrote:

...

"Undelivered Mail Returned to Sender" -> Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda"

Difficult to do anything about this without a gdb backtrace. Can you reproduce it by manually running dovecot-lda? If not, getting a core dump would be the next best way to get a backtrace. http://dovecot.org/bugreport.html

Have you seen such a permission problem cause a crash? If something crashes rather than logging "Permission denied", I'm definitely interested in fixing it.

Anyway, since this happens only sometimes, not always, it can't really be a permission problem.

On 4.10.2010, at 20.18, fakessh wrote:

this is probably a problem of permission

service auth { unix_listener /var/spool/postfix/private/auth { group = postfix mode = 0666 user = postfix } unix_listener auth-master { mode = 0666 } vsz_limit = 256 }

Le lundi 04 octobre 2010 à 21:13 +0200, spamvoll@googlemail.com a écrit :

...

im not sure if i enabled core dumps

im using Centos 5.5 i echoed "'DAEMON_COREFILE_LIMIT="unlimited"' >> /etc/sysconfig/dovecot" and restarted dovecot no core files

but i dont have any killed messages in my logs anyway like: dovecot: Apr 23 11:16:05 Error: child 86116 (imap) killed with signal 11 only " (Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda")"

2010/10/4 spamvoll@googlemail.com:

...

hard to reproduce.. happens only with some mail.

with mail_debug = yes i cant see any errors, but the mail get delivered 2 times ?

Oct 4 20:17:39 imap dovecot: lda(woelky@example.com): sieve: msgid=374f9ab6-5153-47b3-bbfc-cde0705c55d1@email.android.com: stored mail into mailbox 'INBOX' Oct 4 20:17:39 imap postfix/pipe[14085]: 411E6F5572: to=woelky@example.com, relay=dovecot, delay=1, delays=0.53/0.03/0/0.44, dsn=2.0.0, status=sent (delivered via dovecot service) Oct 4 20:17:40 imap postfix/pipe[14082]: 411E6F5572: to=malte.woelky@example.com, relay=dovecot, delay=1.1, delays=0.53/0.02/0/0.53, dsn=5.3.0, status=bounced (Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda")

2010/10/4 Timo Sirainen tss@iki.fi:

...

On Mon, 2010-10-04 at 14:36 +0200, spamvoll@googlemail.com wrote:

...

"Undelivered Mail Returned to Sender" -> Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda"

Difficult to do anything about this without a gdb backtrace. Can you reproduce it by manually running dovecot-lda? If not, getting a core dump would be the next best way to get a backtrace. http://dovecot.org/bugreport.html

-- http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x092164A7

gpg --keyserver pgp.mit.edu --recv-key 092164A7

Am 04.10.2010 16:41, schrieb Timo Sirainen:

On Mon, 2010-10-04 at 14:36 +0200, spamvoll@googlemail.com wrote:

...

"Undelivered Mail Returned to Sender" -> Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda"

Difficult to do anything about this without a gdb backtrace. Can you reproduce it by manually running dovecot-lda? If not, getting a core dump would be the next best way to get a backtrace. http://dovecot.org/bugreport.html

Hi,

I become aware this error today too. I checked my log, first time it ocourrs was Sep. 28, should be with version 2.0.3.

Every time a core file is generated at ~account. (maildir) Can I do something with that file?

After postfix is killed it starts again and delivers this message. I think the message was still queued by postfix. Additionally the sender gets a non delivery.

regards

Mit freundlichen Gruessen --- Burckhard Schmidt

Stephan Bosch writes:

On 10/04/2010 09:38 PM, Schmidt wrote:

...

Am 04.10.2010 16:41, schrieb Timo Sirainen:

...

On Mon, 2010-10-04 at 14:36 +0200, spamvoll <at> googlemail.com wrote:

...

"Undelivered Mail Returned to Sender" -> Command died with signal 11: "/usr/libexec/dovecot/dovecot-lda"

Difficult to do anything about this without a gdb backtrace. Can you reproduce it by manually running dovecot-lda? If not, getting a core dump would be the next best way to get a backtrace. http://dovecot.org/bugreport.html

Hi,

I become aware this error today too. I checked my log, first time it ocourrs was Sep. 28, should be with version 2.0.3.

Every time a core file is generated at ~account. (maildir) Can I do something with that file?

After postfix is killed it starts again and delivers this message. I think the message was still queued by postfix. Additionally the sender gets a non delivery.

regards

Please make a gdb backtrace as described here:

http://www.dovecot.org/bugreport.htm

Regards,

Stephan

This has happened to me too. After a bit of investigation I believe I found what causes it, and I have successfully managed to reproduce it.

This seems to happen when dovecot-lda attempts to deliver a mail to multiple aliases resolving to the same user. For example:

# /etc/aliases xiiph: xiiph@example.com admin: xiiph@example.com

Sending a mail to both xiiph and admin (for example as to and cc) will result in one mail being delivered, and one mail bounced with signal 11 from dovecot-lda.

The server behaves as I would want it to, only deliver one mail and not two to the same user, but I would wish for it to silently ignore the error and suppress the warning mail, as it may confuse senders.

/X

Timo Sirainen writes:

I think that simply causes two dovecot-ldas to be executed at the same time. And that probably causes a race condition that causes the crash. Did you already try patched with http://hg.dovecot.org/dovecot-2.0/raw-rev/e2f9baa436f2 ?

I updated the server to 2.0.6 today, which I noticed already contained the above mentioned patch. Unfortunatenly, it did not solve the problem :(

Is there any workaround I could apply in the mean time, in order to prevent a bounce?

Best regards

/X

Timo Sirainen writes:

Run lda.sh instead: Hopefully with the ulimit -c you'll also get a core dump? gdb backtrace from that would be very helpful then.

I applied your suggested changes and made the shell script above. I replicated the problem and made a gdb backtrace.

In order to avoid "flooding" I uploaded the gdb backtrace. You can find it here: http://bit.ly/9WqDkM

Hope this helps!

Regards,

X

Timo Sirainen writes:

On 26.10.2010, at 9.38, Xiiph wrote:

...

In order to avoid "flooding" I uploaded the gdb backtrace. You can find it here: http://bit.ly/9WqDkM

That looks exactly like the problem that was fixed in 2.0.6. I don't think

you're using 2.0.6's

dovecot-lda. That backtrace couldn't happen with it.

Doh! You are 100% correct, I had compiled 2.0.6 but had not executed the final install command. While waiting for a good time to take the production server down, I forgot all about it.

Thanks a lot for making me aware of this, and for trying to solve a non-existent issue!

Best regards,

X