Stop appending test names reported to Perfherder with -cold and add the pageload type to extra options instead
Categories
(Testing :: Performance, enhancement, P1)
Tracking
(firefox79 fixed)
Tracking | Status | |
---|---|---|
firefox79 | --- | fixed |
People
(Reporter: sparky, Assigned: alexandrui)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
In browsertime tests, we should start using much shorter test names than the raptor-webext variants. We are doing well here so far, but one thing we should do as soon as possible is changing how the test names get changed depending on the pageload type.
For instance, when we run a cold pageload test on amazon, the test name is transformed to amazon-cold, and the warm variant stays as amazon. Instead, we should be adding the cold
and warm
specifiers to the extra options and leaving the test name alone.
:bebe, do you know someone that could take on this work? I think we will have to coordinate it with some dashboard changes since the cold pageload test names will change.
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Updated•5 years ago
|
Assignee | ||
Comment 1•5 years ago
|
||
Sparky, mind if you link here as see also the bug(s) that are adding -cold to test names?
Reporter | ||
Comment 2•5 years ago
|
||
:alexandrui, we add -cold
to the test name in a couple places, we need to make sure that we only modify the browsertime tests and not the existing raptor-webext tests:
(1) https://dxr.mozilla.org/mozilla-central/source/testing/raptor/raptor/manifest.py#426
(2) https://dxr.mozilla.org/mozilla-central/source/taskcluster/taskgraph/transforms/raptor.py#115
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Comment 3•5 years ago
•
|
||
so far I am getting
Traceback (most recent call last):
File "/home/alexandru.ionescu/workspace/mozilla-central/taskcluster/mach_commands.py", line 367, in show_taskgraph
tg = getattr(tgg, graph_attr)
File "/home/alexandru.ionescu/workspace/mozilla-central/taskcluster/taskgraph/generator.py", line 153, in full_task_graph
return self._run_until('full_task_graph')
File "/home/alexandru.ionescu/workspace/mozilla-central/taskcluster/taskgraph/generator.py", line 359, in _run_until
k, v = next(self._run)
File "/home/alexandru.ionescu/workspace/mozilla-central/taskcluster/taskgraph/generator.py", line 273, in _run
raise Exception("duplicate tasks with label " + task.label)
Exception: duplicate tasks with label test-windows7-32/opt-browsertime-tp6-profiling-firefox-amazon-e10s
When we get -cold
out of browsertime tasks, they become their warm versions in terms of naming. We have to have a unique naming for each config version of the test. How do we deal with that?
I am going to submit to phab the current patch and talk there about it.
Assignee | ||
Comment 4•5 years ago
|
||
Reporter | ||
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Updated•5 years ago
|
Assignee | ||
Comment 5•4 years ago
|
||
The changes are already made in
testing/raptor/raptor/manifest.py
. The only thing missing now is adding cold into the extra options.
Sparky, can I kindly ask you to take over this? Doesn't look like a time consuming task and I see that you have more context than me o checking the changes are proper.
Assignee | ||
Updated•4 years ago
|
Assignee | ||
Comment 6•4 years ago
|
||
Nevermind, I'll do it. I added a comment here: https://phabricator.services.mozilla.com/D73035#2389697 Could you please respond?
Comment 7•4 years ago
|
||
This isn't related to the task names, it's about the test names that are reported to Perfherder.
Assignee | ||
Updated•4 years ago
|
Updated•4 years ago
|
Comment 9•4 years ago
|
||
bugherder |
Reporter | ||
Updated•4 years ago
|
Description
•