[1.x] videoroom: remote publisher doesn't release RTP/RTCP ports about janus-gateway

[atoppi]

Hi @boddob thanks for reporting.
Could you try the attached patch? It's on top of current master but it's just the removal of one line, it should not be difficult to adapt it to your current commit.

remove_extra_ref_vr_remote_pub.txt

[boddob]

@atoppi thanks for sharing this. I'll give it a try and get back!

[boddob]

@atoppi I gave this a try and it works as expected now. I'm seeing that the sockets are released now. Thanks! Is there something more to test here to validate the change?

Mar 19 10:46:48.477174 : [plugins/janus_videoroom.c:janus_videoroom_process_synchronous_request:7202:init] 0xffff6000f840 (1)
Mar 19 10:46:48.477174 : [plugins/janus_videoroom.c:janus_videoroom_process_synchronous_request:7221:increase] 0xffff6000f840 (2)
Mar 19 10:46:48.477174 : [plugins/janus_videoroom.c:janus_videoroom_process_synchronous_request:7221:increase] 0xffff6000f840 (3)
Mar 19 10:46:48.480002 : [plugins/janus_videoroom.c:janus_videoroom_process_synchronous_request:7313:increase] 0xffff6000f840 (4)
Mar 19 10:46:48.480002 : [plugins/janus_videoroom.c:janus_videoroom_remote_publisher_thread:12835:increase] 0xffff6000f840 (5)
Mar 19 10:46:48.982088 : [plugins/janus_videoroom.c:janus_videoroom_handler:9473:increase] 0xffff6000f840 (6)
Mar 19 10:46:49.004447 : [plugins/janus_videoroom.c:janus_videoroom_handler:9809:decrease] 0xffff6000f840 (5)
Mar 19 10:48:52.771590 : [plugins/janus_videoroom.c:janus_videoroom_publisher_stream_destroy:2344:decrease] 0xffff6000f840 (4)
Mar 19 10:48:52.771590 : [plugins/janus_videoroom.c:janus_videoroom_publisher_stream_destroy:2344:decrease] 0xffff6000f840 (3)
Mar 19 10:48:52.771590 : [plugins/janus_videoroom.c:janus_videoroom_publisher_dereference:2387:decrease] 0xffff6000f840 (2)
Mar 19 10:48:52.771590 : [plugins/janus_videoroom.c:janus_videoroom_publisher_destroy:2429:decrease] 0xffff6000f840 (1)
Mar 19 10:48:52.771590 : [plugins/janus_videoroom.c:janus_videoroom_remote_publisher_thread:13130:decrease] 0xffff6000f840 (0)

[atoppi]

I guess due to the nature of the fix a quick test is enough but I'll let @lminiero reply in case he feels more testing is needed.

[lminiero]

If you can test it a few more days it will help, just to be sure. But I agree this is 99.99999% likely the cause of the issue.

[boddob]

@atoppi @lminiero thanks, I'll test for some more time and get back.

[atoppi]

@boddob the patch has been merged, thanks for testing.

[boddob]

@atoppi sorry for not getting back in time. I've been running this under heavy loads, and I've not faced any issues.

[zevarito]

I've deployed 2 servers with this patch, one about a week ago, and seems running fine, metrics doesn't show up improvements or regressions compared with the rest of the servers so far. Yesterday I've deployed a second one, that unfortunately crashed about 10hs later. Core dump core.hloop 2.1 was somehow truncated and I am not able to debug and post results. Just wanted to share this in case someone else is also facing a similar issue, for now I would prefer to no update the reset of the servers until have a better insight about the issue.

[boddob]

@zevarito I tested this patch for a week with the latest janus-gateway code at the time and hadn't seen any crash. For production, we have this running for over a month, but patched on an older janus-gateway version (1200, commit : e3a2aea). Will update here if I see any issues.