767348 - Memory consumption shows an ascending trend on Nightly/Win 6.1 x64

Status: RESOLVED INCOMPLETE

(Mozilla QA Graveyard :: Mozmill Tests, defect)

Description

[Mihaela Velimiroviciu (:mihaelav)]

There are 2 memory spikes on June 14 and 15 and an ascending memory consumption trend afterwards on Win7_64bit

Win 6.1 x64
/testTabView_OpenNewTab/test1.js->testOpenNewTab
        average resident  average explicit  max resident  max explicit
13 Jun:      119                79               136          98 
14 Jun:      145               107               175          139

/testTabbedBrowsing_OpenNewTab/test1.js->testOpenNewTab
        average resident  average explicit  max resident  max explicit
13 Jun:      141                99               166          128
14 Jun:      148               108               195          161
Reports:
- http://mozmill-ci.blargon7.com/#/endurance/report/e67171ea696231bb192f56561505d77f (June 13)
- http://mozmill-ci.blargon7.com/#/endurance/report/e67171ea696231bb192f5656151043ff (June 14)

Win 7x64 overall memory increase:
Explicit memory (MB): Maximum: 128 / Average: 75 -> Maximum: 161 / Average: 83
Resident memory (MB): Maximum: 166 / Average: 114 -> Maximum: 195 / Average: 121

Comment 1

[u279076]

Mihaela, whenever you file these bugs, can you also investigate the landings on mozilla-central for this time, providing the hg pushlog? It would be useful to have that information first hand.

Comment 2

[Mihaela Velimiroviciu (:mihaelav)]

(In reply to Anthony Hughes, Mozilla QA (irc: ashughes) from comment #1)
> Mihaela, whenever you file these bugs, can you also investigate the landings
> on mozilla-central for this time, providing the hg pushlog? It would be
> useful to have that information first hand.

Sorry I missed that. Pushlog between June 13 and June 14:
http://hg.mozilla.org/mozilla-central/pushloghtml?startdate=2012-06-13&enddate=2012-06-14

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Mihaela, for a real pushlog please never use dates. Instead use the changesets of the given builds.

Comment 4

[Mihaela Velimiroviciu (:mihaelav)]

(In reply to Henrik Skupin (:whimboo) from comment #3)
> Mihaela, for a real pushlog please never use dates. Instead use the
> changesets of the given builds.

http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=964b11fea7f1&tochange=880f30ecdf6b

Comment 5

[u279076]

Purely speculative, but it looks like SQLite was upgraded in this window (Bug 764243). Perhaps they shipped a memory regression?

Comment 6

[Dave Hunt [:davehunt] [he/him] ⌚BST]

I'm currently still bisecting for bug 764823 due to VM issues. Is there any chance someone can take the investigation of this issue?

Comment 8

[Mihaela Velimiroviciu (:mihaelav)]

After the spikes from June 14 and 15, the memory started to maintain high values after June 16. I'm thinking that the regression may be between June 16 and June 19 (there are no results available between those dates).

Pushlog (June 16-19): 
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=4e3362864fbd&tochange=373e6f9264e6

Comment 9

[Dave Hunt [:davehunt] [he/him] ⌚BST]

It looks like this has dropped back off a little. Mihaela: Would you be able to investigate this further so we can find out what caused the regression? It would involve running the endurance tests on the affected platform for the nightly builds, and then potentially building Firefox to determine the specific change that caused this regression.

Comment 10

[Dave Hunt [:davehunt] [he/him] ⌚BST]

As Mihaela is on PTO, could you pick this up Virgil or Ioana? Thanks.

Comment 11

[Dave Hunt [:davehunt] [he/him] ⌚BST]

I'm going to close this issue as INCOMPLETE. The memory levels dropped considerably after this date, and tinderbox builds are no longer available, so narrowing this down could be a lot of effort for relatively little reward. Now that we have the ability to run tinderbox builds through Mozmill CI we should be much quicker to diagnose such potential memory regressions.