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

Issue 686741 link

Starred by 5 users

Issue metadata

Status: Duplicate
Merged: issue 609137
Owner:
Closed: Apr 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Windows
Pri: 1
Type: Bug-Regression



Sign in to add a comment

Memory Leak From iFrame Navigation

Reported by garystim...@gmail.com, Jan 30 2017

Issue description

UserAgent: Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/56.0.2924.76 Safari/537.36

Steps to reproduce the problem:
1. Please see JSFiddle: https://jsfiddle.net/yrbd0mtw/2/
2. Before clicking the Load Frame Button take a heap snap shot.
3. Click the Load Frame button
4. Periodically take a heap snapshot and observe memory increases continually.

What is the expected behavior?
Memory should be freed as the iframe unloads one page and loads in another like it would with top frame navigation.

What went wrong?
Many of our customer who run our LABVANTAGE product are finding the browser progressively slows and in some cases crashes. On investigation it is found that navigation in iframes leaves all the objects from the iframe still in memory after the iframe is navigated away from. The JS Fiddle https://jsfiddle.net/yrbd0mtw/2/ shows the issue.

Did this work before? Yes 48

Chrome version: 56.0.2924.76  Channel: stable
OS Version: 6.1 (Windows 7, Windows Server 2008 R2)
Flash Version: 

It seems to have improved recently with Chrome 56. However, the problem is still there and is a serious concern for our customers.
This may be related to:
https://bugs.chromium.org/p/chromium/issues/detail?id=529941
and 
https://bugs.chromium.org/p/chromium/issues/detail?id=609137
 
Labels: Prestable-56.0.2924.76 Needs-Triage-M56
Customers upgraded to 56.0.2924.76 beta to try and see if memory problem improves. However they have reported memory still being consumed as they navigate around and are getting the "Snap" error with "Chrome has run out of memory". See attached screenshot.
Thanks,
Gary
memory.png
10.6 KB View Download
Components: Blink>MemoryAllocator>GarbageCollection
Can someone from oilpan take a look at this? As mentioned in the description, it seems similar to  issue 529941  and  issue 609137 
Any update on this? Our customers are chasing for a solution as Chrome is still growing to over 200 - 500 mb in minutes of navigating around and is either making Chrome unresponsive and slow or completely crashing the browser with a out or memory error. 
Thanks,
Gary
Labels: Needs-Bisect
Can someone from the QA team do a bisect? The description says it worked in 48.

Comment 6 by bashi@chromium.org, Feb 16 2017

Owner: pfeldman@chromium.org
Status: Assigned (was: Unconfirmed)
I observed memory growth only when I used DevTools. Probably leaks happen when we use DevTools?

pfeldman@: could you triage this?
Owner: bashi@chromium.org
Original report suggests that there is a memory leak in the browser (without DevTools being opened) and it needs to be fixed.

Leaks w/ devtools opened are considered separately and are of a lower priority - DevTools would retain handlers to some of the obects (console, heap profiler), etc. That's expected and is not observable by the end users.

Comment 8 by bashi@chromium.org, Feb 16 2017

Cc: pfeldman@chromium.org
The repro steps actually use DevTools. If that's an expected behavior, should we close this as WontFix as I didn't observe leaks w/o DevTools?
Labels: -Pri-2 Needs-Feedback Pri-1
No, I think there is an issue here not related to devtools. Maybe just use similar repro steps to  Issue 609137  which uses task manager for a gross heap size estimate.

Do you think this could be a re-occurrence of  Issue 529941 ?

I'm tentatively bumping the priority of this up because it looks like a regression.

Comment 10 by bashi@chromium.org, Feb 16 2017

Chatted offline with dominicc@. I used task manager to observe memory growth and couldn't observe leaks (remained < ~260MB for minutes).

#4: Could you try again on Canary without using DevTools?
Huge thanks for your help on this issue.

I did some investigation this morning and as the issue thread points out it may be more down to the dev tools showing the memory leak. 

However, as the customers don't run dev tools and as they initially reported Chrome slows down and become unresponsive and sometimes crashes, there may be something else at play.

Please see my own testing from this morning and I will now ask the customers experiencing the problem to repeat this process where they are finding the issue. I will also obtain all their system information and ask them to disable any extensions.

Thanks again and I will let you know the outcome of their tests.
Chrome Memory Analysis.pdf
470 KB Download
Cc: krajshree@chromium.org
Unable to reproduce the issue on Win-10 using chrome reported version #56.0.2924.76, latest stable #56.0.2924.87 and latest canary #58.0.3018.0.

Attached a screen cast for reference.

Following are the steps followed to reproduce the issue.
------------
1. Navigated to URL: https://jsfiddle.net/yrbd0mtw/2/
2. Clicked the Load Frame button
3. Observed that the memory increased from 40 MB till 105 MB and never touched 200 MB. It fluctuated at 120 MB and below.

garystimson@ - Could you please check this issue on latest stable #56.0.2924.87 by creating a new profile without any apps and extensions and please let us know if the issue still persist or not.

Thanks...!!
686741.mp4
4.2 MB View Download
Please see my previous comment. I am unable to reproduce with no extensions and no dev tools. Looks like the memory leak I was observing was with dev tools. However our customers are still seeing the problem using latest stable release without dev tools. They have many corporate extensions in place they cannot disable and they are currently working with their IT department to disable the extensions and retest. I will update this thread when we know more.
Thanks,
Gary
Our client is still seeing this issue on their corporate installations of Chrome 56. Even using the JSFiddle above they see the memory constantly increase. I have passed this issue on to their internal engineer to see if they can work with you to reproduce.
Thanks,
Gary
Hi All,

As Gary wrote, we see this issue on our corporate installation of Chrome, version: 56.0.2924.87.
Windows 7 Enterprise, 64-bit, 8 GB RAM, CPU i5-4300U.

Steps I followed to reproduce the issue:
1. Navigate to https://jsfiddle.net/8svao0gp/1/
2. Click the Load Frame button.
3. Observe memory usage on Google Chrome Task Manager.
Results:
After an hour of refreshing the page, memory increased from approximately 40 MB to 187 MB.

Best,
Marek
Labels: -Needs-Bisect -Needs-Triage-M56 M-58
Unable to provide bisect information as this issue is very inconsistent to reproduce.

Could anyone from Blink>MemoryAllocator>GarbageCollection team please have a look into this issue.

Thanks...!!
I am also experiencing this issue. I have a web page that periodically sets the src of an iframe to a different url every 10 seconds.

Using Task Manager, I can clearly see a memory spike every 10 seconds when new content is loaded. The memory usage never goes back down.

This does not seem to happen in Edge.
I forgot to mention. 

I am using Chrome version  57.0.2987.98 (64-bit)
Windows 10 Enterprise, 16GB Ram
Cc: kochi@chromium.org dominicc@chromium.org
Is this still happening (please make sure you don't open any DevTools and don't have extensions)? If not, we may want to mark this as WontFix.
If I read Gary's comment on #13 correctly, without opening dev tools
this exact issue doesn't happen, but something is still leaking memory...
One of the suspect is  issue 609137  (as in comment #9).

Maybe it's worth keep this open for tracking his customer issue to be
resolved.  Merging to 609137 might be another option.

Comment 22 by bashi@chromium.org, Apr 11 2017

Mergedinto: 609137
Status: Duplicate (was: Assigned)
Let me merge this to  issue 609137 . Keeping an issue which no one can repro doesn't seem to be a good thing. Rather, we should file a new bug or re-open the issue as needed.

Please feel free to re-open if the above justification doesn't make sense.

Sign in to add a comment