New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 635033 link

Starred by 2 users

Issue metadata

Status: Verified
Owner:
Closed: Aug 2016
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Linux , Windows , Chrome , Mac
Pri: 2
Type: Bug



Sign in to add a comment

Apprtc calls not getting established due to failed ICE connection state

Project Member Reported by srcv@chromium.org, Aug 5 2016

Issue description

OS: Chrome OS and Linux

Linux:
Chrome Version : 52.0.2743.82 stable
Network: Wired

Chrome OS:
Chrome Version : 54.0.2819.0 / 8676.0.0 dev
Network: Google Guest

What steps will reproduce the problem?
Start a two-way apprtc call by joining the same room from both Linux and Chrome OS using https://appr.tc/r/roomname


Expected result:
The call should be connected successfully, with no audio/video freeze and ICE state breakage

Actual:
1. Apprtc call does not get connected on both sides: Linux and Chrome OS
2. Js logs show: 
"37.262: ICE connection state changed to: failed
apprtc.debug.js:4084 37.265: Callstats status: success msg: autoFabricSetupFailed sent to the backend."

Note:
a)This issue is observed across different networks(Google Guest/Google-A/Wired)
b) This issue is not reproducing consistently (observed 50% of the times)




 

Comment 1 by srcv@chromium.org, Aug 5 2016

Js Logs and webrtc-internals-dump from Linux and Chrome OS:
https://pantheon.corp.google.com/storage/browser/chromiumos-test-logs/bugfiles/cr/635033/
Screenshot 2016-08-05 at 12.15.09 PM.png
855 KB View Download
Labels: OS-Mac OS-Windows
ICE state disconnected happens in apprtc calls b/w Mac & CrOS, and Win & CrOS
JS Console logs from Mac - https://x20.corp.google.com/users/sr/srnarayanan/www/no_crawl/crbug635033

Mac & Win:
Chrome Version : 54.0.2820.0 Canary
Network: Google-A
Labels: Needs-Feedback
This could be a local network issue, hard to know since in the logs it just looks like its not able to connect with the provided candidates.

Could you try to disable ipv6 on both sides and see if the issue remains? https://appr.tc/?ipv6=false

Comment 4 by srcv@chromium.org, Aug 10 2016

This issue was not observed after disabling ipv6 on both sides.

As mentioned in initial bug report, ICE state disconnections were seen only 50% of the times and could not be reproduced consistently.

Comment 5 by srcv@chromium.org, Aug 10 2016

Labels: -Needs-Feedback
ICE state dropped in an ipv6-disabled apprtc call session b/w Linux (wired network) and CrOS. 
Happened 3/3 times that we tried 

JS logs, webrtc-internal dump, and testrtc report here - https://x20.corp.google.com/users/sr/srnarayanan/www/no_crawl/crbug635033/Logs%20from%20Linux%20(wired%20network)
Labels: Needs-Feedback
Owner: srcv@chromium.org
Status: Assigned (was: Untriaged)
Just to be clear, when you write wired, do you mean both machines are connected using an ethernet cable? That's the only way we can rule out the Wifi network behaving badly (it could also the wifi driver/hardware on the CrOS device).

I think it might be something to do with relay candidates not being generated for the callee side. When you try to repro, can you try to figure out if it correlates to which machine initiate the call? E.g. does it happen every time CrOS starts the call first etc.


Owner: jansson@chromium.org
Actually there is nothing you can do, I spoke to the backend team and this is most likely a turn server issue. Will keep you updated.
Owner: srcv@chromium.org
srcv@ could you try this again? The backend team have fixed the issue that hopefully caused this.

Comment 11 by srcv@chromium.org, Aug 17 2016

Labels: -Needs-Feedback
Owner: jansson@chromium.org
Status: Verified (was: Assigned)
No ICE state disconnections were observed during recent AppRTC calls (from 08/15-08/17/2016).


Sign in to add a comment