Closed Bug 829932 Opened 7 years ago Closed 2 years ago

Regularly notify the top users on https://secure.pub.build.mozilla.org/builddata/reports/reportor/daily/highscores/highscores.html of TryServer best practices

Categories

(Release Engineering :: General, defect)

defect
Not set

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: emorley, Unassigned)

References

Details

(Whiteboard: [capacity])

Having looked at http://people.mozilla.org/~catlee/highscores/highscores.html on and off for the last few months, it's clear that there is always a minority of people that use an order of magnitude more machine time than everyone else.

For example, in the last 7 days:

User	                                        Hours	Jobs
tvyas@mozilla.com........................	1,562	3,365
gszorc@mozilla.com.......................	1,247	2,650
VYV03354@nifty.ne.jp.....................	1,078	2,352
cpeterson@mozilla.com....................	694	1,271
alfredkayser@gmail.com...................	661	1,393
Ms2ger@gmail.com.........................	652	1,489
opettay@mozilla.com......................	648	1,299
jcoppeard@mozilla.com....................	639	1,434
dougt@mozilla.com........................	555	1,281
amarchesini@mozilla.com..................	534	1,101
jlebar@mozilla.com.......................	435	880
nfroyd@mozilla.com.......................	411	843
akovalyov@mozilla.com....................	399	947
blassey@mozilla.com......................	397	675
gsharp@mozilla.com.......................	395	910

From an education POV we should:

1) Make sure users are aware that a "-p all -u all -t none" uses a whopping 120 hours of machine time (and >double if they also "-t all").

2) Encourage them to use http://trychooser.pub.build.mozilla.org/ to narrow to one platform/subset of tests, at least until they are sure that their run isn't going to burn everything. In addition, I've seen many people that do a "-p all -u all" run, which is green apart from say an identical mochitest-4 orange on all platforms. Instead of then iterating using "-p <one_platform> -u mochitest-4", then continue to make multiple full platform/tests pushes.

3) Remind people to cancel unwanted runs, eg: https://groups.google.com/d/msg/mozilla.dev.platform/iengKcyD504/DPJeGxOpGw0J

From a technical POV, we've already done:

* Bug 823135 - Show build load in trychooser webpage
* Bug 823907 - Remove the "enabled everything (mozilla-central matching run)" option from TryChooser
* Bug 823711 - Have fewer platforms/unit tests selected by default on the TryChooser page

Other things we could do:

1) Rename the "-p all" and "-u all" options to break the habit of people who've saved mq commits with "-p all -u all" right from the start. Once bug 691177 sticks, perhaps we should use s/all/most/ ?

2) Bug 823713 - The TryChooser page should list the total number of jobs that will be run

3) Bug 674751 - Add Trychooser platform option to chose fastest build
Notified Tanvi of high usage in bug 822367 comment 46.
Depends on: 829933
gps in bug 718066 comment 104 & emk (VYV03354) in bug 827032 comment 17.
Yup. I had to push a lot of Talos jobs to track down a Linux crasher and to examine regressions for bug 718066. Hopefully sanity is near :)
Notified echen in bug 809725 comment 49.
raymond@raysquare.com................	820	1,527
jwalden@mit.edu......................	809	1,721
echen@mozilla.com....................	741	1,606
rlin@mozilla.com.....................	718	1,553
yhuang@mozilla.com...................	714	1,554
mfowler@mozilla.com..................	560	1,160
dholbert@mozilla.com.................	535	1,045
smichaud@pobox.com...................	492	1,216
vyang@mozilla.com....................	476	974
spohl@mozilla.com....................	465	851
saneyuki.snyk@gmail.com..............	464	921
jcoppeard@mozilla.com................	423	847
tvyas@mozilla.com....................	419	855
markcapella@twcny.rr.com.............	417	860
jdrew@mozilla.com....................	372	683
amarchesini-mozilla.com...............	1,824	3,745
raymond-raysquare.com.................	1,214	1,814
echen-mozilla.com.....................	718	1,570
mh-glandium.org.......................	697	1,508
spohl-mozilla.com.....................	666	1,268
Ms2ger-gmail.com......................	636	1,351
vyang-mozilla.com.....................	576	1,129
jdrew-mozilla.com.....................	551	1,104
alfredkayser-gmail.com................	550	1,135
jcoppeard-mozilla.com.................	532	1,136
masayuki-d-toybox.com.................	521	1,163
nmatsakis-mozilla.com.................	494	1,108
mozilladev-lindenbergsoftware.com.....	398	855
tcole-mozilla.com.....................	371	941
kgupta-mozilla.com....................	366	860
There are several of us doing this, and it's ongoing, so unassigning.
Assignee: emorley → nobody
Status: ASSIGNED → NEW
Summary: Regularly notify the top users on http://people.mozilla.org/~catlee/highscores/highscores.html of TryServer best practices → Regularly notify the top users on https://secure.pub.build.mozilla.org/builddata/reports/reportor/daily/highscores/highscores.html of TryServer best practices
Product: mozilla.org → Release Engineering
Found in triage.
Component: Other → Tools
Component: Tools → General
Status: NEW → RESOLVED
Closed: 2 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.