[OpenSIPS-Devel] Erroneous Retransmissions

[Ben Newlin]

Has anyone had a chance to look into this issue? I was able to mitigate by inserting small delays in my test clients, but now I am running into an issue where responses from one OpenSIPS to another are being sent quickly enough to cause this problem. Making changes in my testbed is one thing, but I really don’t want to start injecting delays into my processing script to work around this issue.

The issue is that final responses to a transaction that are received within a few milliseconds of the transaction being sent are not stopping the retransmission timer. The response is handled properly by OpenSIPS, but approximately 500ms later OpenSIPS will retransmit the request anyway. I can reproduce it very regularly in our testbed and it should also be easily reproducible with a simple SIPp scenario.

Trace from the latest occurrence is here: https://pastebin.com/YAXFwNLb

I’d be happy to open an issue on Github as well if that is desired.

Ben Newlin

From: Devel <devel-bounces at lists.opensips.org> on behalf of Ben Newlin <Ben.Newlin at genesys.com>
Reply-To: OpenSIPS devel mailling list <devel at lists.opensips.org>
Date: Thursday, April 25, 2019 at 11:46 AM
To: OpenSIPS devel mailling list <devel at lists.opensips.org>
Subject: Re: [OpenSIPS-Devel] Erroneous Retransmissions

I have found that this issue seems to only occur when the responses are received immediately after the request has been sent; within a few milliseconds. Inserting a delay in the response seems to have made the erroneous retransmissions stop. This does not make it any less of a bug, but does seem to be a workaround.

Ben Newlin

From: Devel <devel-bounces at lists.opensips.org> on behalf of Ben Newlin <Ben.Newlin at genesys.com>
Reply-To: OpenSIPS devel mailling list <devel at lists.opensips.org>
Date: Friday, April 19, 2019 at 8:47 AM
To: OpenSIPS devel mailling list <devel at lists.opensips.org>
Subject: [OpenSIPS-Devel] Erroneous Retransmissions

Resending the issue described below to the devel mailing list as I believe this is a bug, not a configuration issue.

Ben Newlin

From: Users <users-bounces at lists.opensips.org> on behalf of Ben Newlin <Ben.Newlin at genesys.com>
Reply-To: OpenSIPS users mailling list <users at lists.opensips.org>
Date: Monday, April 15, 2019 at 7:20 PM
To: OpenSIPS users mailling list <users at lists.opensips.org>
Subject: Re: [OpenSIPS-Users] Erroneous Retransmissions

Just wanted to bump this to see if anyone has had a chance to take a look. We are continuing to see this failing our tests intermittently. I have a new example here: https://pastebin.com/kQJSGE56.

In this example you can see that the retransmission of the INVITE is occurring not just after the 100 Trying has been received, but also after a 500 final response has already been received and is being processed, but the INVITE is still retransmitted!

Any help on this would be greatly appreciated.

# opensips -V
version: opensips 2.4.5 (x86_64/linux)
flags: STATS: On, DISABLE_NAGLE, USE_MCAST, SHM_MMAP, PKG_MALLOC, F_MALLOC, FAST_LOCK-ADAPTIVE_WAIT
ADAPTIVE_WAIT_LOOPS=1024, MAX_RECV_BUFFER_SIZE 262144, MAX_LISTEN 16, MAX_URI_SIZE 1024, BUF_SIZE 65535
poll method support: poll, epoll, sigio_rt, select.
git revision: d025b4f61
main.c compiled on 21:18:18 Apr 12 2019 with gcc 7

Ben Newlin

From: Users <users-bounces at lists.opensips.org> on behalf of Ben Newlin <Ben.Newlin at genesys.com>
Reply-To: OpenSIPS users mailling list <users at lists.opensips.org>
Date: Thursday, April 4, 2019 at 3:10 PM
To: OpenSIPS users mailling list <users at lists.opensips.org>
Subject: [OpenSIPS-Users] Erroneous Retransmissions

Hello,

We are seeing an issue which is causing our tests to fail intermittently. It appears that sometimes OpenSIPS will retransmit a relayed INVITE request, even though a 100 Trying response has already been received and processed. This only happens intermittently.

I have captured debug logs for the scenario: https://pastebin.com/bHcBJmtv. In the log, you can see the 100 Trying response is received on line 707 and it appears to be matched to the correct transaction on line 729. However then on line 774 you can see the retransmission of the INVITE anyway.

I have verified using tcpdump that the INVITE is being retransmitted; it is not just a log error. And our tests are failing due to the unexpected extra 100 Trying response to the retransmission.

Can anyone explain why this would occur?

Ben Newlin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/devel/attachments/20190508/0f099e91/attachment.html>