| Problem with dgramrcvbuf [message #38161] |
Tue, 24 November 2009 20:06  |
schogge
Messages: 19
Registered: July 2007
|
Junior Member |
|
|
Hi all,
I'm running a machine with two voip applications: asterisk and opensips (formerly known as openser).
The applications where working fine for about one week. Finally the asterisk server was no longer able to communicate to anybody. Neither to external voip-servers nor to the opensips server on the same machine. Restarting asterisk solved the problem and asterisk was working fine again.
I noticed that dgramrcvbuf hit the barrier which probably caused asterisk to fail. To eliminate this error for the future I tripled dgramrcvbuf and other resource limits for this container.
After about two weeks the same error happened again. The load on the machine caused by asterisk and opensips hasn't increased a lot.
Is there a way to figure out what exactly leads to this problem. Dose anybody have any advice how to solve this issue?
Thanks,
schogge
|
|
|
|
| Re: Problem with dgramrcvbuf [message #38165 is a reply to message #38161] |
Wed, 25 November 2009 08:07  |
khorenko
Messages: 533
Registered: January 2006
Location: Moscow, Russia
|
Senior Member |
|
|
Hi schogge,
i'm not sure this is exactly your case, but i'm aware about one recent Asterisk problem - it does not close UDP sockets sometimes, and thus after some time you get too many opened sockets => system becomes inoperable, and yes, asterisk restart helps:
https://issues.asterisk.org/view.php?id=15627
You can check if this is your situation or not by monitoring:
# cat /proc/net/sockstat
Hope that helps.
--
Konstantin
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|
|
| Re: Problem with dgramrcvbuf [message #38169 is a reply to message #38165] |
Wed, 25 November 2009 13:29 |
schogge
Messages: 19
Registered: July 2007
|
Junior Member |
|
|
Hi Konstantin,
this answer is very interesting. Thank you very much.
I'm not a 100 % sure if I have same issue as described in the bug report. But it could be. Unfortunately the bug still exists in current asterisk versions.
I tried to to view sockstat inside my container with
# cat /proc/net/sockstat
but the file seems to be empty. There is no output at all. Is that normal? On the host machine I get this output:
TCP: inuse 261 orphan 0 tw 1 alloc 299 mem 40
UDP: inuse 207 mem 57
RAW: inuse 1
FRAG: inuse 0 memory 0
Rgs,
schogge
[Updated on: Wed, 25 November 2009 13:30] Report message to a moderator |
|
|
|
| Re: Problem with dgramrcvbuf [message #38170 is a reply to message #38169] |
Wed, 25 November 2009 13:51 |
khorenko
Messages: 533
Registered: January 2006
Location: Moscow, Russia
|
Senior Member |
|
|
| Quote: | I tried to to view sockstat inside my container with
# cat /proc/net/sockstat
but the file seems to be empty. There is no output at all. Is that normal?
|
Yes, this is normal.
| Quote: | On the host machine I get this output:
TCP: inuse 261 orphan 0 tw 1 alloc 299 mem 40
UDP: inuse 207 mem 57
RAW: inuse 1
FRAG: inuse 0 memory 0
|
Well, this is not much. i'd suggest just monitor this file, for example, write a script in cron which will dump the info to a file every, say, 15 minutes. And if the node will be again unresponsive from the asterisk's point of view - just check that log: if you see big numbers - that's it.
| Quote: | But it could be. Unfortunately the bug still exists in current asterisk versions.
|
A patch attached to that bug known to fix that problem, so if you get the confirmation that you hit the same problem - it is worth to recompile asterisk from sources with that patch applied.
Hope that helps.
--
Konstantin
If your problem is solved - please, report it!
It's even more important than reporting the problem itself...
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
