Kim Moir [:kmoir] ET Assignee
	Description • 9 years ago

Currently they are set with a global value in config_seta.py in skipconfig_defaults

	Joel Maher ( :jmaher ) (UTC -8)
	Comment 1 • 9 years ago

Vaibhav, would you be interested in taking on this bug?  it is the next usage of SETA- we will use this to tweak windows xp to coalesce more to reduce our usage of windows XP.

	Vaibhav (:vaibhav1994)
	Comment 2 • 9 years ago

Joel, thanks for asking, I am interested to work on it, though may not be able to work on it immediately (so not assigning it to myself).

I wanted to clarify one thing regarding this bug - should we define the skipconfig parameters in SETA API? It would allow us to have access to an API to determine the rate. Or should we define it in config_seta.py?

	Joel Maher ( :jmaher ) (UTC -8)
	Comment 3 • 9 years ago

:kmoir, can you answer vaibhav's question?  I assume you want it in buildbot, but there could be value adding moar data to seta returned via the api.

	Kim Moir [:kmoir] ET Assignee
	Comment 4 • 9 years ago

Joel, so to understand, do you mean that that the json I consume in buildbot would have a skipcount and timeout value for each test on a per platform basis?  So different tests on the same platform might have different intervals to skip them? This is related to https://bugzilla.mozilla.org/show_bug.cgi?id=1219437#c1

	Joel Maher ( :jmaher ) (UTC -8)
	Comment 5 • 9 years ago

yes, this could be related- I am fine adding things wherever it makes the most sense for simplicity and future hacking!

One example would be instead of:
{"jobtypes": {"2015-10-28": ["android-4-3-armv7-api11 mozilla-inbound debug test cppunit", "android-4-3-armv7-api11 mozilla-inbound debug test crashtest-4", ...]}}

we would have:
{"jobtypes": {"2015-10-28": [{"android-4-3-armv7-api11 mozilla-inbound debug test cppunit": {'skipcount': 7, 'timeout': 3600}}, {"android-4-3-armv7-api11 mozilla-inbound debug test crashtest-4": {'skipcount': 7, 'timeout': 3600}},...]}}

you can imagine that we would be able to set an entire platform inside of SETA to have different skip counts, likewise a per test basis (say longer running tests, etc.)

	Kim Moir [:kmoir] ET Assignee
	Comment 6 • 9 years ago

coop asked me to actively work on this today given our very high pending counts so I'm going to work on implementing a per platform values in config_seta.py, and set coallascing on winxp to twice as often as default as per bug 1219437

	Kim Moir [:kmoir] ET Assignee
	Comment 7 • 9 years ago

seta can configured on a per slave platform basis, winxp is skipped at twice the default rate

	Kim Moir [:kmoir] ET Assignee
	Comment 9 • 9 years ago

Looking at the logs after the 9am PT reconfig, this looks like it's working

2015-10-30 08:59:53-0700 [-] tests-fx-team-xp-ix-opt-unittest-7-3600: skipping with 1/7 important changes since only 801/3600s have elapsed
2015-10-30 08:59:53-0700 [-] tests-fx-team-xp-ix-debug-unittest-7-3600: skipping with 2/7 important changes since only 980/3600s have elapsed
2015-10-30 08:59:54-0700 [-] tests-mozilla-inbound-xp-ix-opt-unittest-7-3600: skipping with 3/7 important changes since only 111/3600s have elapsed
2015-10-30 08:59:54-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-7-3600: skipping with 1/7 important changes since only 203/3600s have elapsed
2015-10-30 09:05:04-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-14-7200: skipping with 1/14 important changes since only 65/7200s have elapsed
2015-10-30 09:05:19-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-14-7200: skipping with 1/14 important changes since only 80/7200s have elapsed
2015-10-30 09:05:53-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-14-7200: skipping with 1/14 important changes since only 114/7200s have elapsed
2015-10-30 09:08:39-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-14-7200: skipping with 1/14 important changes since only 280/7200s have elapsed
2015-10-30 09:08:52-0700 [-] tests-mozilla-inbound-xp-ix-debug-unittest-14-7200: skipping with 1/14 important changes since only 293/7200s have elapsed

	Kim Moir [:kmoir] ET Assignee
	Comment 10 • 9 years ago

closing

we might want to change this in the future so we have more granularity with respect to opt or debug platforms, but it is working on slave platforms for now