Hi @rgupta0110 ,
Manage to replicate the bug, now working on the fix. The problem is that
somehow epoll_wait generates EPOLLERR event(strerror(errno)="Bad file
descriptor") so we will try to come with the fix as soon as possible.
---
Reply to this email directly or view it on GitHub:
@rgupta0110 i would be very thankful if you could try this patch[0] and tell me
if it solves your problem. It solved the problem for me.
Regards,
Ionut Ionita
[0] https://gist.github.com/ionutrazvanionita/86efca1c13c6654c2a74
---
Reply to this email directly or view it on GitHub:
Hi Bogdan,
I have uploaded the tcpdump traces for the issue along with core, binaries and
similar log files
http://www88.zippyshare.com/v/kgfvUJ9b/file.html
Opensips IP - 10.205.235.5 (port 5060)
CallManager - 10.205.235.5 (Port 5059)
endpoint - 10.205.203.130
Thanks
Rahul Gupta
---
Reply
Hi Bogdan, this is inside our lab and it won’t be possible to give you the
access. However, I can work with you to put any debug binaries or code changes
and test it out. Let me know if that’s something we can do.
Thanks
Rahul Gupta
From: Bogdan Andrei IANCU [mailto:notificati...@github.com]
Rahul, the missing element for me is how the TCP conn gets terminated while in
the SYN_SENT state (after attempting to connect to the disconnected party).
What is actually happening with the conn while in SYN_SENT ? I get is not
opensips the one destroying itis there some TCP timeout in
@rgupta0110 thank you for your effort. In order to speed up the fixing, is it
possible to get access to an opensips where this can be reproduced (once again,
I couldn't reproduce it on my side) - so I can trigger the testing and have
opensips im my hands to change and debug ? If you can manage
The last logs you provided actually indicate that the poll event was an
EPOLLIN, while that fd is added only for EPOLLOUT :-/ . This is way, even if
your fix hides the effects, it is not dealing with the actual problem.
---
Reply to this email directly or view it on GitHub:
Hi Bogdan, I have reproduced the issue and uploaded the core file along with
opensips bins,libs and other logs at the following location
http://www81.zippyshare.com/v/LdXDN5r6/file.html
Call flow and timestamps :
1) Tue Aug 11 11:36:54 --- created a TCP connection to send out SIP NOTIFY to
Hi Bogdan, I will generate the core file and upload the logs and core soon.
Meanwhile, I have put the following fix which seems working and I haven't
experienced any issues so far. Could you please review it and let me know if
there is any potential problem
Added the following condition in
@rgupta0110 , that may be a simplistic way of fixing the effect, but without
understanding the cause. As I mentioned, in my tests I get a completely
different behavior, so fixing the effects may not bit a generic solution.
Please let me know about the ERROR logs that I asked you to enable and
@rgupta0110 , I did a lot of testing and research on this. Still I was not able
to reproduce the the behavior you describe (at least not the last step). While
I have the conn in SYN_SENT, if (A) conn is expired by opensips, it is simply
removed or if (b) the kernel is timing it out (due low
Server:: OpenSIPS (1.11.3-tls (x86_64/linux))
FYI, its reproducible with 1.11.5 latest src too.
---
Reply to this email directly or view it on GitHub:
https://github.com/OpenSIPS/opensips/issues/591#issuecomment-128704584___
Devel mailing list
@rgupta0110 , which opensips version were you using for this logs ?
---
Reply to this email directly or view it on GitHub:
https://github.com/OpenSIPS/opensips/issues/591#issuecomment-128635411___
Devel mailing list
Devel@lists.opensips.org
Hi Bogdan,
Almost a month ago, I have raised the issue regarding the opensips log file
filling up with the following message
ERROR:core:io_watch_del: BUG - trying to del fd 38 with flags 2 1
Now I can reproduce it every time with the following scenario:
1) UAC and UAS are connected
I have captured logs for the scenario and uploaded to zippyshare . Please let
me know if you want me to upload to anyother location
http://www76.zippyshare.com/v/njlgk4lW/file.html
1) netstat_debug.log --- netstat info for the particular endpoint
(10.205.203.130 in this case)
2)
15 matches
Mail list logo