We're using STABLE CVS-Nv1-0-5-02/24/05 and we've been noticing that
sometimes there's a socket leak on REGISTER SIP messages. We've seen it
happen only on customers using Sipura SPA2100 ATAs.
If I issue a "sip show channels", I see thousands of "zombie
channels".
If I look into the details, that's what I get - actually one single
"sip
show channel <channelID>" returns thousands of these:
* SIP Call
Direction: Incoming
Call-ID: 5c2cf755-ccde1b6c@x.x.x.xq: 520 REGISTER
Our Codec Capability: 12
Non-Codec Capability: 1
Their Codec Capability: 0
Joint Codec Capability: 0
Format unknown
Theoretical Address: x.x.x.x:5060
Received Address: x.x.x.x:5060
NAT Support: RFC3581
Our Tag: 715659627
Their Tag:
SIP User agent:
Need Destroy: 0
Last Message:
Promiscuous Redir: No
Route: N/A
DTMF Mode: rfc2833
The sequence number (ie. 520) increases by 1 every time.
After a while, I run out of files and I have to restart asterisk. I have
temporarily solved the problem by issuing a "ulimit -n 8192" in
safe_asterisk, but that's not a solution, since I will eventually reach
that limit as well. Is there a way to fix this? We're running RHEL4 and
we have about 300 customers registered all the time.
Thank you very much
-Manuel
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://lists.digium.com/pipermail/asterisk-users/attachments/20050307/4207e0eb/attachment.htm
