[asterisk-bugs] [JIRA] (ASTERISK-28992) app_voicemail: Deadlock in ODBC when retrieving file

[Schneur Rosenberg (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-28992?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=252911#comment-252911 ] 

Schneur Rosenberg edited comment on ASTERISK-28992 at 12/3/20 8:08 AM:
-----------------------------------------------------------------------

Hi Sean, I applied the new patch and it does compile and the voicemail still gets stuck but it does get released after a few minutes, so its only semi fixed, but I did lots of testing and and I'm a bit smarter now, you can now reproduce my issue very easily, the system will only crash when the msgnum isnt sequencial, so if I delete a voicemail from the DB without reordering the msgnum I will see a message in the CLI something like "app_voicemail_odbc.c:9146 open_mailbox: Resequencing Mailbox: /var/spool/asterisk/voicemail/default/odbctest/INBOX, expected 3 but found 2 message(s) in box with max threshold of 100." this could go on for a few minutes depending on how many messages I have in the VM box, and while its reordering the msgnum field the channel will hang, prior to the new patch it would remain hung now it does get released but way after I hang up.

During my testing the system just deleted 50 messages one time when I hung up while it was hanging, also it didnt always delete the .lock file when the channel get release and no new VMs could be left until I manually deleted it


was (Author: thesipguy):
Hi Sean, I applied the new patch and it does compile and the voicemail still gets stuck but it does get released after a few minutes, so its only semi fixed, but I did lots of testing and and I'm a bit smarter now, you can now reproduce my issue very easily, the system will only crash when the msgnum isnt sequencial, so if I delete a voicemail from the DB without reordering the msgnum I will see a message in the CLI something like "app_voicemail_odbc.c:9146 open_mailbox: Resequencing Mailbox: /var/spool/asterisk/voicemail/default/odbctest/INBOX, expected 3 but found 2 message(s) in box with max threshold of 100." this could go on for a few minutes depending on how many messages I have in the VM box, and while its reordering the msgnum field the channel will hang, prior to the new patch it would remain hung now it does get released but way after I hang up.

During my testing the system just deleted 50 messages one time when I hung up while it was hanging.

> app_voicemail: Deadlock in ODBC when retrieving file
> ----------------------------------------------------
>
>                 Key: ASTERISK-28992
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-28992
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Applications/app_voicemail/ODBC
>    Affects Versions: 17.5.1
>         Environment: Debian 9
>            Reporter: Schneur Rosenberg
>            Assignee: Unassigned
>              Labels: patch
>         Attachments: 0001-app_voicemail-Prevent-deadlocks-when-out-of-ODBC-dat.patch, locks.zip
>
>
> I'm having a major issue since I upgraded from Asterisk 11 that both voicemail and voicemailmain ocasionally crash, the call remains connected but there is silence, and the channel cannot be killed, hanging up wont help and neither would  "hangup request" and the every few seconds there would be a message in the CLI like WARNING[31353]: chan_sip.c:4410 __sip_autodestruct: Autodestruct on dialog '1216767442-5060-95 at BJC.BGI.A.BA' with owner SIP/9174342412-0000016a in place (Method: BYE). Rescheduling destruction for 10000 ms.
> Usually once it crashes no other users can get into their voicemail boxes and no new voicemails can be left until i restart asterisk



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