[asterisk-bugs] [JIRA] (ASTERISK-24041) deadlock when cdr_pgsql fails to post CDR due to PGSQL connection failure

[Gergely Dömsödi (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-24041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=220502#comment-220502 ] 

Gergely Dömsödi commented on ASTERISK-24041:
--------------------------------------------

Attached all debug files. These where created when the channel got stuck in asterisk.

The hangup from one of the participant comes in at 10:51:37 in the main-cut log. The other files were created at 10:51:55, a little while after the lock happened.

There seems to be no way to remove the stuck channel from asterisk.

Right now, we use this system with the CDR turned off, but if more logs or tests are required, I think I could reproduce the issue by turning it back again.

> deadlock when cdr_pgsql fails to post CDR due to PGSQL connection failure
> -------------------------------------------------------------------------
>
>                 Key: ASTERISK-24041
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-24041
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: CDR/cdr_pgsql
>    Affects Versions: 1.8.28.0
>            Reporter: Gergely Dömsödi
>         Attachments: backtrace-threads.txt, core-channels.txt, core-locks.txt, core-threads.txt, main-cut.log
>
>
> On our production system SIP channels got stuck in asterisk. This happened 1-2 times daily on a system which is barely used. After a lot of trial-and-no-result, I managed to seemingly solve it by turning off the posting of CDRs to PGSQL, hence the assumption that it is somehow related to cdr_pgsql. 
> In the logs, we could see a SIP hangup coming in from one leg of the channel, the UserEvent fireing which we have in the "h" extension, but the asterisk didn't send the BYE event to the other participant.
> Unfortunately I couldn't reproduce the issue on a non-production system, but I managed to get a lot of debug log from the production server.



--
This message was sent by Atlassian JIRA
(v6.2#6252)