New issue
Advanced search Search tips

Issue 725337 link

Starred by 2 users

Issue metadata

Status: Duplicate
Merged: issue 728030
Owner:
Closed: Jun 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Android
Pri: 1
Type: Bug

Blocking:
issue 728452



Sign in to add a comment

ServiceWorker.StartWorker.Status ERROR_START_WORKER_FAILED is increasing on Android

Project Member Reported by shimazu@chromium.org, May 23 2017

Issue description

On Android canary/dev, the ratio of ERROR_START_WORKER_FAILED has increased. 

https://uma.googleplex.com/timeline_v2?sid=04f43fdf4cbf58307f259d832beb83b8

Something introduced between 60.0.3097.0 and 60.0.3096.4 might be causing the error.
The number itself is small, but let's investigate it because the number of unique users has also increased. 
 

Comment 1 Deleted

Comment 2 by horo@chromium.org, May 31 2017

Summary: ServiceWorker.StartWorker.Status ERROR_START_WORKER_FAILED is increasing on Android (was: ServiceWorker.StartWorker.Status ERROR_START_WORKER_FAILED is increasing)

Comment 3 by horo@chromium.org, May 31 2017

https://uma.googleplex.com/timeline_v2?sid=2ded92135073d3af96c225d1631ecd75

The rate of ServiceWorker.StartWorker.Status ERROR_START_WORKER_FAILED on android Canary is 00.96% at May 29.

Comment 4 by falken@chromium.org, May 31 2017

Anecdotally, lately on Chrome Canary my Add to Homescreen apps (Twitter, Facebook) don't really load. It shows the splash screen indefinitely.

Strangely, once I open Chrome and navigate to chrome://serviceworker-internals to try to debug it, and return to the app, it shows the splash screen then starts loading.
I suspected there were any changes around memory, process management or some new experiment and tried to find plausible changes by scanning the log from `git log --format=oneline`, but it failed.

This is the uma split by versions:
https://uma.googleplex.com/p/chrome/timeline_v2/?sid=5327eec6a2b142781d8d196efd29f401
Weirdly, nhiroki@ just hit this on Android Stable 58.

Since we have similar phones, I wonder if it could be limited to this device. If the bug is on Stable I would have expected more bug reports.
> Weirdly, nhiroki@ just hit this on Android Stable 58.

I opened the Twitter Lite from the homescreen on 58.0.3029.83 and its splash screen showed indefinitely. According to the chrome://serviceworker-internals, at the time SW state is...

- Installation Status: Activated
- Running Status: Stopped

The app worked well after terminating it from the task switcher and launching it again from the homescreen.
This is a trace when
1. Open Twitter Lite from the home screen
2. Wait for 20 sec
3. Open Chrome Canary
4. Open chrome://serviceworker-internals
5. Back to Twitter Lite from app history

https://drive.google.com/open?id=0Bwd71LAFvG5ScVVmdHhIY2xIQzQ

This shows that StartWorker is blocked. Long PREPARING_SCRIPT_LOAD starts from sending StartWorker IPC message and finishes at creating URLJob for the main script. 
I suspect the allocated process for the service worker might be sleeping when issuing the StartWorker message and the IPC didn't wake the process up.
Screenshot from 2017-06-01 10:59:19.png
11.4 KB View Download

Comment 9 by kbr@chromium.org, Jun 1 2017

Blocking: 728452
Cc: altimin@chromium.org
This might be related to this: https://codereview.chromium.org/2808843003
Do you think it's relevant to your patch?

What we are now observing is

- when opening a website, swiping chrome canary out from the app history on android and opening chrome again, the page won't be reloaded.
- when opening the Twitter Lite from the home screen, it won't be loaded until launching chrome and opening twitter.com from the chrome. 

Just to be clear, pages that don't use service worker are also failing to load, right?
Status: Started (was: Assigned)
re c#11:
Yes, all pages cannot be loaded after chrome is closed.

I reverted the patch locally and it works. Let me revert the patch.

Comment 13 by horo@chromium.org, Jun 1 2017

 Issue 726901  may be related to this.
I saw "E/cr_ChildProcessService(32184): Service has not been bound with bindToCaller()" error in my device.
Thanks, I could repro the problem with monochrome on 61.0.3116.0, and it seems already solved by https://chromium-review.googlesource.com/c/518289. 
However,  Issue 726901  has started on 60.0.3111. I guess there might be another problem.
Cc: clamy@chromium.org
Labels: Proj-PlzNavigate
Mergedinto: 728030
Status: Duplicate (was: Started)
It's due to the PlzNavigate issue reported on  https://crbug.com/705318#c57  . 

I could confirm it from the result of finch:

60.0.3101.4 (Dev, May 17)
https://uma.googleplex.com/p/chrome/variations/?sid=31ff975825baf80ad0252727dc6d1907

60.0.3096.5 (Dev, May 17)
https://uma.googleplex.com/p/chrome/variations/?sid=50d0391b51e265515952129c4c1ff5e9

60.0.3097.0 (Canary, May 13)
https://uma.googleplex.com/p/chrome/variations/?sid=3bf342cc50cbf13f4aeef3ae984b42ab

60.0.3096.4 (Canary, May 12)
https://uma.googleplex.com/p/chrome/variations/?sid=4651564a0ae9005aab449c3446660879

There are only two CLs in the range which have "PlzNavigate" in the log  (https://chromium.googlesource.com/chromium/src/+log/60.0.3096.4..60.0.3097.0?pretty=fuller&n=10000 ).

* PlzNavigate: Fix wrong "Cache-Control" header. (https://codereview.chromium.org/2863683003)
* Don't enforce X-Frame-Options for downloads. (https://codereview.chromium.org/2874933002)

I don't understand what the changes are. Let me pass the issue to PlzNavigate folks.

Sign in to add a comment