Closed Bug 938464 Opened 11 years ago Closed 6 years ago

google search results page takes 100-500k in sessionstore file

Categories

(Firefox :: Session Restore, defect)

x86
All
defect
Not set
normal

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: tnikkel, Unassigned, NeedInfo)

References

(Blocks 1 open bug)

Details

(Keywords: perf, Whiteboard: [fxperf:p3] p=0)

Using https://addons.mozilla.org/en-US/firefox/addon/about-sessionstore/ to look at my session store file I have a bunch of google search results pages open, they all seem to take between 100-500k, which seems far too much. By far the worst offender in my open tabs.
Status: NEW → RESOLVED
Closed: 11 years ago
Resolution: --- → DUPLICATE
Doesn't really seem like a dupe of that bug to me. This bug is about a specific site causing bloated sessionstore entries. The other bug is about fixing sessionstore in general.
Status: RESOLVED → REOPENED
Resolution: DUPLICATE → ---
Bug 669603 is pretty much about google search storing lots of sessionStorage data which leads to a huge sessionstore.js. There is really no need to track this twice.
Oh, I see. I had just read the bug title and assumed it to be about moving long sessionstore operations off the main thread or async, and was thinking this bug could be about either getting google to not store so much data, or doing something on our end so that we don't have to store so much data.
Can we leaf this bug open, change 'Component' to 'Tech Evangelism', 'Platform' to 'All All' for Google to have an look on and keep in mind the Mozilla try to fix this in Bug 669603 ???
Keywords: perf
OS: Mac OS X → All
Whiteboard: [triage]
uggh. with a 2014-01-23 build, for search tabs less than 24hrs old, for different search strings I'm seeing
700k
221k
85k
221k
Whiteboard: [triage]
Whiteboard: p=0
No longer blocks: fxdesktopbacklog
Flags: firefox-backlog+
Does this still happen for you?  

(Haven't tested myself)
Flags: needinfo?(tnikkel)
Flags: needinfo?(lee.thomson)
(In reply to Wayne Mery (:wsmwk, NI for questions) from comment #8)
> Does this still happen for you?  
> 
> (Haven't tested myself)

Yep, tested in a fresh profile, one google search result tab had 490k.
Flags: needinfo?(tnikkel)
(In reply to Wayne Mery (:wsmwk, NI for questions) from comment #8)
> Does this still happen for you?  
> 
> (Haven't tested myself)

Yes, it appears to still persist.
about:sessionstore no longer shows individual tab info, watching sessionstore size differentials shows;
Single Google search result page: ~200k
Single duck duck go " : ~600 bytes

4 Google: 816,126 Bytes
4 Duck Duck Go: 2,799 Bytes
I'm a bit surprised your numbers are so low - after sorting the about:sessionstore list all large tabs are google searches, one of them 4,5MB, 11 of them more than 1MB in size, so you can probably imagine how large my sessionstore is... This is definitely still happening with 51.0a2(2016-10-21)
After some discussion on IRC, started by a user with the same problem, we found that disabling Google Instant improved the situation by a lot (to ~8k per page, which is still a lot, but at least more manageable). Just dropping this here, in case it might help someone.
I also still see this - 100k-500k per search
Flags: needinfo?(lee.thomson)
I'm getting between (100 K and 6 MB) for 100% of all the "Google Search" tabs listed in about:sessionstore.

"Google Search" is the only site with this problem. All the other tabs are small, around ~4 K. 

The problem occurs with "Google Instant predictions: (●) Never show Instant results" -- Google Instant is *disabled*. I have double checked and there is no possibility that it is enabled. I do not see any instant results, and it is disabled in preferences.

I am seeing the same problem on multiple computers.

I am seeing the same problem on multiple versions of Firefox.

I am seeing the same problem with a fresh profile and complete reinstall of the browser.
I have ~500 tabs with many Google Searches. Due to the large size of my sessionstore, I am seeing this error every time I attempt to exit the browser:

http://i.imgur.com/TVXqW7b.png

If I don't press "continue" then my session becomes corrupted.
Currently, I'm not seeing this issue on my machine. Here is a screen shot from about:sessionstore in FF 54:

http://imgur.com/a/nwOve

The storage from Google search is really small, only around 230 bytes. Can somebody confirm this?
Is Google Instant enabled? (see comment #12 and comment #14)
Flags: needinfo?(nnn_bikiu0707)
I don't think so. [1] shows that Google may have remove the Google Instant feature. In fact, I never changed any settings from Google Search. A while back ago, I did look at about:sessionstore and the storage from Google Search is very large. But now, everything looks fine, I guess.

:dietrich, are you still experiencing this problem?

Also, Brzola, are you experiencing this? What is your Google settings? And it would be really helpful if you can attach your about:support page.

[1]: https://www.theverge.com/2017/7/26/16034844/google-kills-off-instant-search-for-mobile-consistency
Flags: needinfo?(nnn_bikiu0707)
Flags: needinfo?(dietrich)
Flags: needinfo?(bzihrla)
No, I'm not experiencing it - I just had remembered it was about Instant.

It's also possible that people have old tabs from back when Instant was around could still see this problem. It would only go away once the tab was closed (and pushed out of closed-tabs data).

IIRC there was also another separate bug about perf of large sessionStorage data, which might also address that general issue. I can't find the bug right now though.
Flags: needinfo?(dietrich)
(In reply to Dietrich Ayala (:dietrich) from comment #19)
> IIRC there was also another separate bug about perf of large sessionStorage
> data, which might also address that general issue. I can't find the bug
> right now though.

Might that have been bug 669603 ?
Priority: -- → P3
Whiteboard: p=0 → [fxperf] p=0
Much of this comes back to bug 669603 and 669034's observations; we persist storage in session history that chrome doesn't, so they see no reason to stop.  There were several ideas proposed those bugs that would avoid or sidestep the resultant issue; or we could follow chrome and not persist it.  Don't know what edge does.  And don't know what impact this has; it can't be horrible given chrome (and maybe all other browsers) don't persist here.

The spec does not mandate persisting, IIRC
Mike, do you have concrete ideas here given comment #21?
Flags: needinfo?(mdeboer)
Priority: P3 → --
Whiteboard: [fxperf] p=0 → [fxperf:p3] p=0
Nope, I think this might actually be a RESOLVED FIXED, because Tim fixed bug 1362058.
Flags: needinfo?(mdeboer)
(In reply to Mike de Boer [:mikedeboer] from comment #23)
> Nope, I think this might actually be a RESOLVED FIXED, because Tim fixed bug
> 1362058.

Sounds like we can resolve wfm per that bug. Please reopen with more details if you're still seeing significant issues with Firefox >= 59.
Status: REOPENED → RESOLVED
Closed: 11 years ago6 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.