Closed Bug 1188459 Opened 9 years ago Closed 9 years ago

[Flame] Performance has regressed in latest; app start up times have increased, returning from apps is slower

Categories

(Firefox OS Graveyard :: Gaia::System, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

(blocking-b2g:2.5+, b2g-v2.2 unaffected, b2g-master fixed)

VERIFIED FIXED
FxOS-S4 (07Aug)
blocking-b2g 2.5+
Tracking Status
b2g-v2.2 --- unaffected
b2g-master --- fixed

People

(Reporter: onelson, Assigned: etienne)

References

()

Details

(Keywords: regression, smoketest, Whiteboard: [2.5-Daily-Testing])

Attachments

(3 files)

Description:
On the latest Flame build [20150728030208], performance has taken a hit across the device. Apps appear to load slower, return to home slower. Attempting to open Card View will display only a single app (in most cases; observed Dialer + Contacts open simultaneously) when opened with multiple apps, as other app cards will close.

Repro Steps:
1) Update a Flame to 20150728030208
2) Open the Clock app
3) Return to home
4) Open the Calendar app
5) Return to home
6) Open the Settings app
7) Return to home
8) Hold home button to open card view

Actual:
App start up times and return times are slower; only one card is displayed in card view

Expected:
Apps start up and return reasonably; all cards displayed in card view


Environmental Variables:
Device: Flame 2.5 [319mb]
Build ID: 20150728030208
Gaia: 14e32276025b0310d3e89027320cf4b2a24cedfb
Gecko: 33dc8a83cfc0
Gonk: 41d3e221039d1c4486fc13ff26793a7a39226423
Version: 42.0a1 (2.5)
Firmware Version: v18D
User Agent: Mozilla/5.0 (Mobile; rv:42.0) Gecko/42.0 Firefox/42.0


Repro frequency: 5/5
See attached: 
video
logcat
Issue DOES NOT REPRODUCE on 2.2 for flame devices
Results: Apps open and return to home reasonably; card view displays all apps as cards that were open

Device: Flame 2.2
BuildID: 20150728032504
Gaia: e1e6317f17a840b19af9dbb25f5a771d8d9fa161
Gecko: 9b60e57724db
Gonk: bd9cb3af2a0354577a6903917bc826489050b40d
Version: 37.0 (2.2) 
Firmware Version: v18D
User Agent: Mozilla/5.0 (Mobile; rv:37.0) Gecko/37.0 Firefox/37.0
QA Whiteboard: [QAnalyst-Triage?]
Flags: needinfo?(pbylenga)
Memory inflation could have made bug 1172167 have worse results.
Adding smoketest keyword, and reminder that this is tested on a 319mb device.
Keywords: smoketest
[Blocking Requested - why for this release]:
Severe performance regression that fails smoke tests.

Requesting a window.  Also let's check Flame with 512mb
blocking-b2g: --- → 2.5?
QA Whiteboard: [QAnalyst-Triage?]
Flags: needinfo?(pbylenga)
Gregor, can you find someone to assign this to please?  It looks like we have a general performance issue.
Flags: needinfo?(anygregor)
Looking at Raptor, Calendar looks like it has a recent regression.
QA Contact: jmercado
This issue doesn't occur with 512 MB Flame.  Performing the window with 319 MB.

Environmental Variables:
Device: Flame 2.5 KK (512 MB)
BuildID: 20150728072944
Gaia: a8319543833b08e986fa3ed590faeb2624bf1683
Gecko: bc589dd18ad5
Version: 42.0a1 (2.5) 
Firmware Version: v18D-1
User Agent: Mozilla/5.0 (Mobile; rv:42.0) Gecko/42.0 Firefox/42.0
Eli, have you noticed anything on raptor?
Flags: needinfo?(anygregor) → needinfo?(eperelman)
The changes for Bug 1186587 seem to have caused this issue.

Central Regression Window:

Last Working 
Environmental Variables:
Device: Flame 2.5
BuildID: 20150723130445
Gaia: ec2199b324304d3678b6a98a08a31bdc13c9e984
Gecko: cb8bdb8ffaef
Version: 42.0a1 (2.5) 
Firmware Version: v18D-1
User Agent: Mozilla/5.0 (Mobile; rv:42.0) Gecko/42.0 Firefox/42.0

First Broken 
Environmental Variables:
Device: Flame 2.5
BuildID: 20150724020739
Gaia: db299103dd17a873cba4ff88e62ec173bf80fe62
Gecko: b0b3dcfa5557
Version: 42.0a1 (2.5) 
Firmware Version: v18D-1
User Agent: Mozilla/5.0 (Mobile; rv:42.0) Gecko/42.0 Firefox/42.0

Last Working gaia / First Broken gecko - Issue does NOT reproduce
Gaia: ec2199b324304d3678b6a98a08a31bdc13c9e984
Gecko: b0b3dcfa5557

First Broken gaia / Last Working gecko - Issue DOES occur
Gaia: db299103dd17a873cba4ff88e62ec173bf80fe62
Gecko: cb8bdb8ffaef

Gaia Pushlog: https://github.com/mozilla-b2g/gaia/compare/ec2199b324304d3678b6a98a08a31bdc13c9e984...db299103dd17a873cba4ff88e62ec173bf80fe62
Blocks: 1186587
QA Whiteboard: [QAnalyst-Triage?]
Flags: needinfo?(pbylenga)
Flags: needinfo?(ktucker)
Raptor does appear to show some sort of regression among several apps:

http://raptor.mozilla.org/#/dashboard/script/apps.js?device=flame-kk&branch=master&memory=319&series=coldlaunch.visuallyLoaded&from=1436994146000&to=1438117533063

It's tough to pinpoint anything specific, since last week bitbar devices were down and we had a lot of trouble getting the new Raptor CLI deployed to automation. But at the least Raptor seems to agree about the increase.
Flags: needinfo?(eperelman)
NI on Etienne, seems the commit for bug 1186587 caused a performance regression, can we get a backout?
Flags: needinfo?(pbylenga) → needinfo?(etienne)
QA Whiteboard: [QAnalyst-Triage?] → [QAnalyst-Triage+]
Flags: needinfo?(ktucker)
Reverting commit 62d9420fc0915ab11ad5d23e105c929d4b5afae6 (the one for bug 1186587) fixes this issue.
Flags: needinfo?(ktucker)
Need info Naoki for possible backout of this smoke test blocker.
Flags: needinfo?(ktucker) → needinfo?(nhirata.bugzilla)
kgrandon is looking at backing out and making sure things are green before merging.
Flags: needinfo?(nhirata.bugzilla)
Flags: needinfo?(kevingrandon)
Bug 1186587 has been backed out, so this should now be fixed.

https://github.com/mozilla-b2g/gaia/commit/a317a3edc00eea564d39bd89d032a76546477cd7
Status: NEW → RESOLVED
Closed: 9 years ago
Flags: needinfo?(kevingrandon)
QA Contact: jmercado → kevingrandon
Resolution: --- → FIXED
Keywords: verifyme
Verifying fix on master for flame devices based on comment 15

Device: Flame 2.5
BuildID: 20150730030209
Gaia: bf8565e0c3ad216ccb3f109c17f8a2eb2c42f6b8
Gecko: 62469b20ec84
Gonk: 41d3e221039d1c4486fc13ff26793a7a39226423
Version: 42.0a1 (2.5) 
Firmware Version: v18D
User Agent: Mozilla/5.0 (Mobile; rv:42.0) Gecko/42.0 Firefox/42.0

Results: Start up times for apps on flame have recovered


Worth noting that performance on flame still isn't great, but that may also be our own memory being low on our test devices. However, the immediate performance impact observed when this bug was filed has appeared to dissipate due to the backout. Also, bug 1172167 is still being observed and that issue remains open: cards will not remain open often when there are multiple apps being brought to card view (commonly all cards but the active app will close when card view is opened).
Status: RESOLVED → VERIFIED
Keywords: verifyme
Moving the bug to the component where the regression came from.
blocking-b2g: 2.5? → 2.5+
Component: General → Gaia::System
Assignee: nobody → etienne
Target Milestone: --- → FxOS-S4 (07Aug)
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: