964323 - Autophone - s1s2 - Regression in "time to throbber start" and "time to throbber stop" on January 24, 2014

Status: RESOLVED WORKSFORME

(Firefox for Android Graveyard :: General, defect)

Description

[Geoff Brown [:gbrown]]

http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstart/local-blank/2014-01-20/2014-01-27/notcached/noerrorbars/standarderror

and

http://phonedash.mozilla.org/#/org.mozilla.fennec/throbberstop/local-blank/2014-01-20/2014-01-27/notcached/noerrorbars/standarderror

show regressions on January 24 for many devices.

For example, consider the time to throbber start time for 38_aa_3c_1e_a8_ee_samsung-gs3:

2014-01-23 18:34:50 (mozilla-central/rev/9d650c07b547): 1228±15
2014-01-24 04:11:50 (mozilla-central/rev/bfe4ed6d47ce): 1831±55

Comment 1

[:Margaret Leibovic]

Brian, is this the regression you were looking into? Does the fix to delay when we fetch snippets fix this?

Comment 2

[Brian Nicholson (:bnicholson)]

Yeah, wlach verified that disabling snippets fixed this regression. Bug 964511 should help by moving this out of the startup path.

Comment 3

[:Margaret Leibovic]

I can take this so that we don't lose track of it.

wesj and I were talking about how this bug possibly relates to bug 966580, as well as to the more extreme eideticker startup regression that we suspect was caused by bug 964511 and friends.

Comment 4

[:Margaret Leibovic]

According to mfinkle, phonedash was changed recently to reduce noise and read data from faster internal memory, so the baseline has changed on these graphs.

However, this issue does appear to be resolved, and we did land bug 969511 to address the snippets problem, so I don't think there's anything actionable for us to do here.