All users were logged out of Bugzilla on October 13th, 2018

The "All Trees" view should display data for all trees, even the ones OrangeFactor doesn't know about

RESOLVED FIXED

Status

P2
normal
RESOLVED FIXED
3 years ago
3 years ago

People

(Reporter: emorley, Assigned: emorley)

Tracking

Details

Attachments

(3 attachments)

(Assignee)

Description

3 years ago
Most of the time, the trees that are of interest are either the "trunk" trees, or else an individual sheriffed release tree.

However if we stop commenting on bugs, it will be much harder to answer the question of "did this intermittent first occur on the try repo or one of the twigs?".

I don't think it's desirable to add every single repo to the list that OrangeFactor knows about, since otherwise:
(a) it will get out of date fast
(b) they'll get included in the "push count" for "All Trees", and so skew the orangefactor score, given that most of the "other" trees are not sheriffed, so will look like they have very low scores

As such, I think the best way to do this is just to make "All trees" not pass any filter on the 'tree' param to ElasticSearch (so we get all classifications submitted by Treeherder, regardless of tree), but to leave the hg pushlog query as-is, so we only count pushes from the main trees (for a start, there isn't an easy way to query the pushlog for "all trees" without an explicit list).

This should be fine, since it only impacts the "all trees" view, and to be honest, the orangefactor score has little meaning there as it is anyway - it's main use is to see occurrence history for a particular intermittent.
(Assignee)

Comment 1

3 years ago
(In reply to Ed Morley [:emorley] from comment #0)
> I don't think it's desirable to add every single repo to the list that
> OrangeFactor knows about, since otherwise:
> (a) it will get out of date fast
> (b) they'll get included in the "push count" for "All Trees", and so skew
> the orangefactor score, given that most of the "other" trees are not
> sheriffed, so will look like they have very low scores

(c) It will clutter up the tree selection menu
(d) It will slow down the pushlog polling part of the query
(Assignee)

Comment 2

3 years ago
Created attachment 8655690 [details] [diff] [review]
Part 1: Rename 'tree' to 'elasticsearch_tree_arg'
Attachment #8655690 - Flags: review?(jgriffin)
(Assignee)

Updated

3 years ago
Attachment #8655690 - Attachment description: Rename 'tree' to 'elasticsearch_tree_arg' → Part 1: Rename 'tree' to 'elasticsearch_tree_arg'
(Assignee)

Comment 3

3 years ago
Created attachment 8655691 [details] [diff] [review]
Part 2: Rename 'unquoted_tree' to 'pushlog_tree_list'
Attachment #8655691 - Flags: review?(jgriffin)
(Assignee)

Comment 4

3 years ago
Created attachment 8655692 [details] [diff] [review]
Part 3: Fetch all data from ES for "All Trees", not just from known trees
Attachment #8655692 - Flags: review?(jgriffin)
Attachment #8655690 - Flags: review?(jgriffin) → review+
Attachment #8655691 - Flags: review?(jgriffin) → review+
Comment on attachment 8655692 [details] [diff] [review]
Part 3: Fetch all data from ES for "All Trees", not just from known trees

Review of attachment 8655692 [details] [diff] [review]:
-----------------------------------------------------------------

Thanks, and +1 for the micro patches.
Attachment #8655692 - Flags: review?(jgriffin) → review+
(Assignee)

Comment 6

3 years ago
https://hg.mozilla.org/automation/orangefactor/rev/2be82bf89bf4
https://hg.mozilla.org/automation/orangefactor/rev/98ecd7abb52d
https://hg.mozilla.org/automation/orangefactor/rev/45bf68cceef6

And deployed.

@sheriffs: Choosing "All" from the tree selection menu on OrangeFactor now returns failure classifications for all trees, not just the ones that OrangeFactor knows about - eg you can now see "Try" here (even though it's not in the tree menu):
http://brasstacks.mozilla.com/orangefactor/?display=Bug&bugid=950636&startday=2015-08-26&endday=2015-09-02&tree=all

...this will allow choosing "All" to see whether an intermittent failure started on Try / a twig, before then occurring on the main trunk branches.
Status: ASSIGNED → RESOLVED
Last Resolved: 3 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.