Swarming page causes Chrome to be unusable |
||||||||||||||||||||||
Issue descriptionSite: https://chromium-swarm.appspot.com/task?id=4261753546188d10&refresh=10&show_raw=1 [Occurred again later with https://chromium-swarm.appspot.com/task?id=427039f13cdc0f10&refresh=10&show_raw=1] Symptoms: Chrome was very barely usable. [10s to switch tabs, etc.]. I was unable to take a heap dump. I was unable to take a trace via chrome://tracing. The renderer for the site and the browser were both spinning at ~150% cores. I was able to take activity monitor samples of both of the processes. The browser process IO thread seems to be spinning at 100% performing extension + networking related work. The renderer process seems to be non-stop requesting resources. Not sure what we can do without a reliable repro. This might fall into the category of an extension DOSing the browser?
,
Jan 18
(5 days ago)
+ellyjones in case she has any ideas about what to do with the problem I posted in c#0 +piman in case he has any insight into the problems with c#1 [unfortunately I don't have any insight about the renderers, I had to kill them before the machine locked up. This canary chrome is still unusable now, minutes afterwards, in case there's any insight we can gain from an unusable chrome]
,
Jan 18
(5 days ago)
busy IO thread will prevent the GPU work from making progress.
,
Today
(16 hours ago)
#2: I don't really know what to add - it looks like we're just doing a ton of work. Does disabling extensions (any specific extension or all of them) ameliorate this?
,
Today
(13 hours ago)
My device from c#1 was still spinning this morning after many hours of up time. I took a sample but forgot to save it -- it looked very similar to the one in c#2. This suggest some type of mojo-related live-lock on the IO thread. Unfortunately, I don't think we have enough info to debug. :( +rockot to look at the sample in c#2 in case he has any insight into how Mojo could live-lock.
,
Today
(13 hours ago)
That stack looks like the stack you see when a child process dies. Is it possible that something is causing us to rapidly spawn new processes that then die immediately?
,
Today
(11 hours ago)
I was able to repro the issue again with steps in opening comment. After I killed the browser process, the renderers have continued !! to spin at 100% CPU on their IO threads, consuming ~10GB each. :(
|
||||||||||||||||||||||
►
Sign in to add a comment |
||||||||||||||||||||||
Comment 1 by erikc...@chromium.org
, Jan 18 (5 days ago)158 KB
158 KB View Download