725701 - Memory usage spike detected in Mozmill Endurance tests for Win 5.1.2600 x86

Status: RESOLVED DUPLICATE of bug 724553

(Mozilla QA Graveyard :: Mozmill Tests, defect)

Description

[Mihaela Velimiroviciu (:mihaelav)]

There is a memory spike of about 30MB between Feb 6 and Feb 8 on the Win 5.1.2600 x86 nighlies.

Chart:
http://mozmill-release.blargon7.com/#/endurance/charts?branch=13.0&platform=All&from=2012-01-31&to=2012-03-13

Report from 2012-02-06:
http://mozmill-release.blargon7.com/#/endurance/report/a3c1112b91483a199fcd4a91cf04dee1

Report from 2012-02-07:
http://mozmill-release.blargon7.com/#/endurance/report/6c25050e24006b3ca556dcd53704cd8a

Report from 2012-02-08:
http://mozmill-release.blargon7.com/#/endurance/report/6c25050e24006b3ca556dcd5370c3fbf

Pushlog between Feb 6 and Feb 8:
http://hg.mozilla.org/mozilla-central/pushloghtml?startdate=2%2F6%2F2012&enddate=2%2F8%2F2012

Note: Win 6.0.6002 x86 also has higher memory usage in this period, but the difference is not so big (less than 10MB)

Comment 1

[u279076]

I'm not sure a spike of 30MB is enough to worry about on WinXP, especially since looking at the historical data it tends to fluctuate day to day. 

Looking at the data since the merge I can see the normal range is between 115MB and 140MB. I This does beg the question, what is a realistic threshold for error? 

Using the Explicit Memory numbers:
 * Average Max Memory: 131 MB
 * Max Max Memory: 143 MB
 * Min Max Memory: 115 MB
 * Std Deviation: 28 MB (21% of Average)

Given that, I think an acceptable norm is a 20% deviation. In the case of Firefox 13 on Windows XP that translates to ~130MB - ~194MB.

Dave, what do you think?

Comment 2

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Yeah, this looks like quite a spike. It coincides with new tests landing but on inspection these new tests are not causing the spike. It looks to me like the private browsing test is causing this spike again.

Comment 3

[u279076]

Should we dupe to bug 724677?

Comment 4

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Fine by me.

Comment 6

[u279076]

Sorry, wrong bug.