[asterisk-bugs] [JIRA] (ASTERISK-24938) ARI Snoop Channel eventually consumes entire CPU
Rusty Newton (JIRA)
noreply at issues.asterisk.org
Fri Apr 3 16:29:32 CDT 2015
[ https://issues.asterisk.org/jira/browse/ASTERISK-24938?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Rusty Newton updated ASTERISK-24938:
------------------------------------
Assignee: George Ladoff
Status: Waiting for Feedback (was: Triage)
Press 'Send Back' or 'Enter Feedback' when responding to make sure we see the response. Thanks!
> ARI Snoop Channel eventually consumes entire CPU
> ------------------------------------------------
>
> Key: ASTERISK-24938
> URL: https://issues.asterisk.org/jira/browse/ASTERISK-24938
> Project: Asterisk
> Issue Type: Bug
> Security Level: None
> Components: Applications/app_stasis, Resources/res_ari_channels
> Affects Versions: 13.2.0, 13.3.0
> Environment: Ubuntu Linux 14.04.2, Asterisk built from source, Realtime with MySQL
> Reporter: George Ladoff
> Assignee: George Ladoff
>
> When I create a Snoop Channel with ARI, the thread immediately begins to consume an additional 2-3% CPU per minute. By about 40 minutes, an entire 3Ghz CPU is consumed.
> To recreate this, simply have any channel in Stasis then create a Snoop channel from the original channel. CPU usage will start escalating right away.
> I'm specifically creating channels with Spy Out but I've tried both directions and Whisper with the same result.
> This happens every time a Snoop channel is created. The Stasis application and all of the channels are still usable, but only until all of the CPU is consumed. Then of course nothing works.
> Once the Snoop channel is deleted, the usage goes immediately back to normal (sustained 1-2% per thread).
> I've tried snooping channels that are using a variety of codecs: ulaw, gsm, g.729. All have the same result.
> This happens whether the Snoop channel is then bridged or recorded or neither.
> I've tried with both Timerfd and pthread.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
More information about the asterisk-bugs
mailing list