switch caroline-tot-chrome-pfq-informational builder to cave |
||||||||
Issue descriptioncaroline DUT availability in the lab is particularly scarce. The workaround in issue 835949 won't last long. It is easier to get cave: https://bugs.chromium.org/p/chromium/issues/detail?id=835949#c9 So consider 1) switch the caroline builder to use cave. 2) require less DUTs to work on the tests. 3) increase suite timeouts 4) run less tests 5) get away from pool based scheduling. It is not clear why informational builders need a pool, of this size and then not automatically maintained.
,
May 7 2018
Didn't figure out where to go. Next gardener might contact ihf@ for code points.
,
May 7 2018
,
May 8 2018
Addressing 2) less DUTs here https://chromium-review.googlesource.com/#/c/chromiumos/chromite/+/1049487
,
May 8 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/chromite/+/8cf7ea75d5ffc48222bec84907256377b6f2e587 commit 8cf7ea75d5ffc48222bec84907256377b6f2e587 Author: Ilja H. Friedel <ihf@chromium.org> Date: Tue May 08 23:39:11 2018 chromeos_config: drop informational min_duts to 1. Allow the continuous pools to do their job even with 1 DUT. Currently the workload looks like a single DUT should be able to cover all testing. So no need to provision 4 DUTs if one can do. But even if one DUT were not enough to cover informational testing, it is better in times when working DUTs are scarce in the lab for the suite to sometimes time out (infra flake) than to refuse to start testing (permanent infra failure). BUG=chromium:836360 TEST=look at caroline graphs https://viceroy.corp.google.com/chromeos/dut_utilization?board=caroline&model=&pool=managed%3Acontinuous&status=Running&topstreams=40&duration=108858&mdb_role=chrome-infra&refresh=-1&utc_end=1525119484 Maybe, maybe minnie needs a minimum of 2, but I think the suite grabs all idle DUTs in the pool anyways (which would be 6 here). https://viceroy.corp.google.com/chromeos/dut_utilization?board=veyron_minnie&model=&pool=managed%3Acontinuous&status=Running&topstreams=40&duration=54429&mdb_role=chrome-infra&refresh=-1&utc_end=1525313968 Change-Id: I5864e1c41d356b4acf5d6722cf86c08fe258b18b Reviewed-on: https://chromium-review.googlesource.com/1049487 Trybot-Ready: Ilja H. Friedel <ihf@chromium.org> Reviewed-by: Pohsien Wang <pwang@chromium.org> Reviewed-by: Ilja H. Friedel <ihf@chromium.org> Tested-by: Ilja H. Friedel <ihf@chromium.org> [modify] https://crrev.com/8cf7ea75d5ffc48222bec84907256377b6f2e587/cbuildbot/config_dump.json [modify] https://crrev.com/8cf7ea75d5ffc48222bec84907256377b6f2e587/cbuildbot/chromeos_config.py
,
May 8 2018
5) should address the issue. What's the reason prevents us from merging the continuous pool with pool:suite?
,
May 9 2018
continuous pool has a bit better service, as it takes devices from suite when they break.
,
May 9 2018
At least that is the theory. One could merge continuous pool into another pool, like bvt or cq. But this is a more long term effort (ideally by writing a new scheduler). I will look again if we can switch to cave. And if we can reduce some continuous pool sizes a little.
,
May 14 2018
forward to next gardener
,
Jun 15 2018
,
Jun 15 2018
|
||||||||
►
Sign in to add a comment |
||||||||
Comment 1 by tbarzic@chromium.org
, Apr 30 2018Owner: warx@chromium.org