1363440 - Gmail tab throbber keeps spinning on latest Nightly

Status: RESOLVED WORKSFORME

(Core :: General, defect)

Description

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

User agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:55.0) Gecko/20100101 Firefox/55.0
BuildId: 20170508030204

This morning I got prompted to update Nightly and restart which I did. After restart, gmail tab throbber keeps spinning.

On the bottom status bar, I don't see any notifications like "waiting for...."/"transferring data....."/. 

This is a terrible experience on a top site.

Comment 1

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

I also noticed in the task manager that the CPU is spinning at 30% in Nightly, even when I am not actively using Nightly Firefox.

Comment 2

[Chris Peterson [:cpeterson]]

Mike, do you know if there have been any recent changes to the tab spinners?

I noticed that the tab spinner never stops on some Google Docs in recent Nightly builds. I haven't seen this problem with Gmail however.

Comment 3

[Mike Conley (:mconley) (:⚙️)]

Are either of you able to reproduce reliably? Are either of you willing to use mozregression to help us identify the regressing changeset?

Comment 4

[Mike Conley (:mconley) (:⚙️)]

(I ask, because I'm having no luck reproducing the issue on today's Nightly with the steps provided)

Comment 5

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Mike and I talked about this on IRC. We are planning to disable tcp_fastopen on my Nightly, restart it, try to repro the same gmail throbber spinning problem a few times with gmail tab in the background, foreground, as an app tab. 

I noticed that downloading a file from https://www.activestate.com/activepython/downloads/thank-you?dl=http://downloads.activestate.com/ActivePython/releases/2.7.13.2714/ActivePython-2.7.13.2714-win64-x64-402182.exe was taking 32 minutes on my Nightly instance was taking 32 minutes. I don't think it is a problem with my network per se. It could be a networking issue on my Nightly Firefox instance (tcp fast open?) or a problem with the ActiveState.com servers.

Comment 6

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

So I was unable to find a correlation between the tcp_fastopen pref and the 30% CPU spike I am noticing on the more recent builds.

Comment 7

[Mike Conley (:mconley) (:⚙️)]

Hm - the excessive CPU could be a separate issue. Any permanent tab throbbers yet?

Comment 8

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

I ran mozregression-script.py --bad 2017-05-09 --good 2017-05-01

32:38.82 INFO: Narrowed inbound regression window from [82c2d17e, 96605941] (3 revisions) to [d7e40bb8, 96605941] (2 revisions) (~1 steps left)
32:38.82 INFO: Oh noes, no (more) inbound revisions :(
32:38.82 INFO: Last good revision: d7e40bb852ea047a0e5f530bcdc04d29a1765001
32:38.82 INFO: First bad revision: 96605941c0021795376d9c6ec1e458de2fac329e
32:38.82 INFO: Pushlog:
https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=d7e40bb852ea047a0e5f530bcdc04d29a1765001&tochange=96605941c0021795376d9c6ec1e458de2fac329e

Mozregression seems to point it to the backout of Bug 1356448 - Enable "GPU" process on Windows software backends by default on Nightly

With every build that mozregression ran, I would open the following tabs: gmail, amazon, facebook, linked, cnn, youtube, google.com and login to gmail,Fb,LI, click around tabs for a bit (5 seconds). I would then watch for tab throbbers spinning and CPU usage. On the good builds, CPU usage would go down from 60-80% to 5-6% and stay at ~30-40% on the bad builds. I am unable to repro the tab throbber problem yet.

Comment 9

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

(In reply to Mike Conley (:mconley) - At work week, slow to respond from comment #7)
> Hm - the excessive CPU could be a separate issue. Any permanent tab
> throbbers yet?

Filed Bug 1363559 for the CPU usage problem.

Comment 10

[Chris Peterson [:cpeterson]]

(In reply to Mike Conley (:mconley) - At work week, slow to respond from comment #3)
> Are either of you able to reproduce reliably? Are either of you willing to
> use mozregression to help us identify the regressing changeset?

Sorry. I don't have STR or specific docs that always demonstrate the problem. In fact, I haven't seen the problem today.

Comment 11

[Mike Conley (:mconley) (:⚙️)]

(In reply to Ritu Kothari (:ritu) from comment #9)
> (In reply to Mike Conley (:mconley) - At work week, slow to respond from
> comment #7)
> > Hm - the excessive CPU could be a separate issue. Any permanent tab
> > throbbers yet?
> 
> Filed Bug 1363559 for the CPU usage problem.

Have you seen any tab perma-throbbers since disabling tcp_fastopen?

Comment 12

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Nope, I have not been able to repro it with/without fast open. I'll resolve this as works for me for now. If I encounter it again, I'll try to attach a gecko profile.

Comment 13

[Mike Conley (:mconley) (:⚙️)]

(In reply to Ritu Kothari (:ritu) from comment #12)
> Nope, I have not been able to repro it with/without fast open. I'll resolve
> this as works for me for now. If I encounter it again, I'll try to attach a
> gecko profile.

Well, hold on - is Dragana aware of this bug? Hey Dragana - it sounds like tcp_fastopen might contribute to this bug... is this a known issue?

Comment 14

[Dragana Damjanovic [:dragana]]

(In reply to Mike Conley (:mconley) - At work week, slow to respond from comment #13)
> (In reply to Ritu Kothari (:ritu) from comment #12)
> > Nope, I have not been able to repro it with/without fast open. I'll resolve
> > this as works for me for now. If I encounter it again, I'll try to attach a
> > gecko profile.
> 
> Well, hold on - is Dragana aware of this bug? Hey Dragana - it sounds like
> tcp_fastopen might contribute to this bug... is this a known issue?

There was an issue with FastOpen and it was fixed in bug 1362821 and also verified in bug 1363223. It think it landed in Nightly on 05/08 or 05/09
if someone notice this afer that Nightly please let me know.