[OpenSIPS-Users] [Crash Report] Weird crash with drouting/tls_mgm/usrloc/db_postgresql
volga629 at networklab.ca
Tue Aug 11 17:26:36 EST 2020
That commit is helped a lot to identify the problem if opensips can't connect to database, but it not improve stability.
Stability is mix protocols and underlining library, opensips just use api to connect, so all mention tweaks are still need it.
From: "Adrien Martin" <a.martin at alphalink.fr>
To: "volga629" <volga629 at networklab.ca>, "OpenSIPS users mailling list" <users at lists.opensips.org>
Sent: Tuesday, August 11, 2020 1:02:24 PM
Subject: Re: [OpenSIPS-Users] [Crash Report] Weird crash with drouting/tls_mgm/usrloc/db_postgresql
I tried to reduce the number of workers to 1 with Opensips 3.1 (can not change TCP workers with 2.4+) and there is another crash about tls_domain that i have to investigate :)
Our problems are somewhat different though, in this one there is no load/no calls, just some registrations (less than 5 UAC) and some drouting probing UDP and TLS.
As #1579 was closed, did the commit https://github.com/OpenSIPS/opensips/commit/c1403a1d9bee2254a84a866352266a41d9ff93bc fixed your problem or did you just tweak the workers/sysctl.conf/postgresql.conf?
Le 11/08/2020 à 14:34, Slava Bendersky a écrit :
> Hello Everyone,
> You need tweak sysctl and postgresql.conf to allow opensips connect properly to remote postgresql.
> In opensips I can't pass for any transport udp_worker tcp_worker more then 10. Other wise it starts segfaulting. One opensips open from 150-170 connections to database under load when cluster of 3 nodes you need postgresql at least 650 connection limit. If you use TCP or WSS or TLS you need increase tcp protocol buffer on protocol level other wise it will be bottleneck for database and it will slow down all.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Users