1095116 - Similar jobs list should always list the current job

Status: RESOLVED INCOMPLETE

(Tree Management :: Treeherder, defect, P3)

Description

[Steve Fink [:sfink] [:s:]]

Another surprise - when using the similar jobs tab from a job from several days ago, I was surprised to not find it anywhere in the list. I expected the similar jobs list to always have the selected job, together with some on each side of it (if any are available). Probably fewer on the "newer" side.

I guess this also means adding an "add in newer jobs" button in addition to (what should be) the "add in older jobs" button.

Comment 1

[Ed Morley [:emorley]]

The similar jobs feature is a bit flaky at the moment, and not used by many people. With hyper-chunking I think its usability is going to get worse (due to varying chunk names), so it's probably not worth investing too much PR/review time into it at the moment, given how short-staffed we are.

Comment 2

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

Sheriffs almost exclusively use the "Similar Jobs" quickly check the same job (build, test chunk) across many pushes - it's faster than filtering treeherder's main view by the generated job hash. This would still work with hyper-chunking, won't it? If I remember correct, I saw a try push with it enabled that "M(Core)" (?)

Investigating intermittent failures on older pushes is really tedious if one has to let treeherder's "Similar jobs"' load jobs several times to get the current job (jobs not being ordered by push order is the other issue).

Comment 3

[David Burns :automatedtester]

(In reply to Sebastian Hengst [:aryx][:archaeopteryx] (needinfo on intermittent or backout) from comment #2)
> Sheriffs almost exclusively use the "Similar Jobs" quickly check the same
> job (build, test chunk) across many pushes - it's faster than filtering
> treeherder's main view by the generated job hash. This would still work with
> hyper-chunking, won't it? If I remember correct, I saw a try push with it
> enabled that "M(Core)" (?)
> 
> Investigating intermittent failures on older pushes is really tedious if one
> has to let treeherder's "Similar jobs"' load jobs several times to get the
> current job (jobs not being ordered by push order is the other issue).

Since you can't get to the logs without having to do clicking how is this process better than filtering jobs using the treeherder and looking at intermittents that way?

Comment 4

[David Burns :automatedtester]

closing as no further info was given