Have you checked if there is traffic on looping on your OpenSIPS ? maybe
due bad routing, OpenSIPS is keep sending to itself some traffic,
causing the loops and the load.
OpenSIPS Founder and Developer
OpenSIPS Summit 2019
On 03/07/2019 03:17 PM, Pasan Meemaduma wrote:
Thanks for the reply. writing inline replies
"Using t_on_reply() is harmless as it simply set a holder in the msg/transaction,
nothing more. And a locally generated reply does NOT trigger on_reply_route - this route
is triggered only by *RECEIVED* replies."
My problem was that server received 407 replies from its own source ip :/. I
couldn't figure out what triggered it. Anyhow issue resolved after I removed
fix_dialog_route() function which causes BYE to send to a private ip of a
NATed device. I have fix_nated_contact() before create_dialog() function but
dialog data seems to have unmodified contact data which I believe is fixed in
2.4 as you mentioned.
"So the bottom problem for you is why the processes stuck ? do they stuck like
consuming cpu (doing some crazy looping) or blocking in some I/O (no CPU usage) ?"
All traffic processing workers were stuck in a loop and consume 100% CPU. I
couldn't get much debug info like running opensipsctl trap when issue occured :/
Anyhow now I'm planning to upgrade thee system to 2.4 this year which has
support till 2021.
my sip cluster is now stable without fix_dialog_route() function.
On 02/13/2019 09:36 AM, Pasan Meemaduma via Users wrote:
How's it possible locally generated replied triggering on_reply_route ? I ran
in to issue where all opensips process gets stuck in processing same call
replies and causing other traffic to get drop.
<info> /usr/sbin/opensips: Call: Reply from a NAT endpoint - S=408
D=Request Timeout F=sip:xxx@xxx T=sip:yyy
yy@x IP=a.b.c.d ID=asgasgasgas
Request process by opensips before this is an ACK request belong to the call
where I don't think It'll expect a reply. Could it be an issue If I call
t_on_reply on an ACK msg ?
I'm trying to figure out where the bug in my opensips routing script. It
causes all sip listerner processes to get stuck in a loop causing to generate
above message. IP a.b.c.d is the sip server IP which confuse me as locally
generated replies shouldn't trigger on_reply_route as per docs.
Any clue is welcome.
I'm using opensips 2.3.6
Users mailing list
Users mailing list