As the subject sez., rollovers just stop working.
Detailed scenario - We use the lcr module for outbound calls through our (multiple) gateway providers. For various reasons, the providers frequently reject the call, and we roll the call over to the next provider (next_gw(), etc., etc.) The lcr module, however, stop rolling the calls over after a couple of thousand rejections. As best as I can tell, next_gw() fails, and the rejection just gets passed back to the CPE.
Any ideas? Is there some sort of counter? Some sort of memory limit that we are running into?
For what it is worth, we process a couple of thousand calls an hour, and the box that OpenSER is running on is a dual Opteron with 3GB, etc., etc.
Cheers
p.s. One additional factoid. We log to acc/mysql, and query the acc table to get CDRs. The query runs once every 30 seconds or so. We can't turn this off for any extended period, but have done so once next_gw() starts failing. It doesn't seem to make a difference.
Mahesh Paolini-Subramanya writes:
For various reasons, the providers frequently reject the call, and we roll the call over to the next provider (next_gw(), etc., etc.) The lcr module, however, stop rolling the calls over after a couple of thousand rejections.
in 1.0.1, load_gws()/next_gw() has no "memory" between calls. of course, it is possible that there is a memory leak in lcr module (or some other place), which makes lcr functions fail. do you get any error message in syslog?
-- juha
Nope. No error messages. That is what makes this whole thing so strange...
cheers
Juha Heinanen wrote:
Mahesh Paolini-Subramanya writes:
For various reasons, the providers frequently reject the call, and we roll the call over to the next provider (next_gw(), etc., etc.) The lcr module, however, stop rolling the calls over after a couple of thousand rejections.
in 1.0.1, load_gws()/next_gw() has no "memory" between calls. of course, it is possible that there is a memory leak in lcr module (or some other place), which makes lcr functions fail. do you get any error message in syslog?
-- juha
Sooo, been tracing the logs, and I've noticed that when the problems start occuring (i.e., the outbound call doesn't rollover when I get a rejection from our gateway provider), I see the following in syslog
ERROR: get_out_socket: no socket found forward_req: ERROR: cannot forward to af 2, proto 1 no corresponding listening socket
What socket d'you think it is referring to?
cheers
Mahesh Paolini-Subramanya wrote:
Nope. No error messages. That is what makes this whole thing so strange...
cheers
Juha Heinanen wrote:
Mahesh Paolini-Subramanya writes:
For various reasons, the providers frequently reject the > call,
and we roll the call over to the next provider (next_gw(), etc., etc.)
The lcr module, however, stop rolling the calls over after a
couple of > thousand rejections. in 1.0.1, load_gws()/next_gw() has no "memory" between calls. of course, it is possible that there is a memory leak in lcr module (or some other place), which makes lcr functions fail. do you get any error message in syslog?
-- juha
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Mahesh Paolini-Subramanya writes:
ERROR: get_out_socket: no socket found forward_req: ERROR: cannot forward to af 2, proto 1 no corresponding listening socket
What socket d'you think it is referring to?
af 2 is inet address and proto 1 is udp. is your proxy multi-homed?
-- juha
-
Juha Heinanen -
Mahesh Paolini-Subramanya