<jmaher> armenzg: I looked into the bc6 failure (browser_crashedTabs.js) <jmaher> guess what...it normally takes 80+ seconds to run, lets add 20-30% overhead and we cross the 90 second timeout ... <jmaher> https://dxr.mozilla.org/mozilla-central/source/browser/components/sessionstore/test/browser_crashedTabs.js#6 <jmaher> armenzg: default timeout is 45 seconds for browser-chrome, ^ makes it 90, lets change that to requestLongerTimeout(3) <jmaher> sadly ehsan just bumped the timeout 5 weeks ago ... <jmaher> ehsan: just that we need to do it again because docker runs tests ~20% slower ... <jmaher> ehsan: it wasn't sad that you bumped it, just sad that we need to do it again <ehsan> we should revise the harness timeout <jmaher> ehsan: the default timeout for browser-chrome is 45 <jmaher> 45 seconds <ehsan> I think we should bump it up to 1:30 mins <jmaher> so 90 seconds, and remove a lot of the extra multipliers? <ehsan> I'd leave the multipliers in <ehsan> they won't really hurt us <jmaher> alright <ehsan> unless if one of those tests actually times out in which case we're wasting a few mins <ehsan> machine time $$$ <<< human $$$ <jmaher> well, we can justify it "because docker takes longer to run, lets do this across the board" <ehsan> jmaher: smart plan :D <ehsan> r-me <ehsan> err <ehsan> = :D <jmaher> ehsan: well, we don't want a job sitting for 20 minutes, then timing out- ideally we can fail faster <jmaher> ehsan: for browser-chrome only or mochitest/chrome also? <jmaher> armenzg: here is where the default is set: https://dxr.mozilla.org/mozilla-central/source/testing/mochitest/browser-test.js#3 <ehsan> jmaher: I don't think we have a lot of tests with multipliers > 4 <ehsan> 4*90 << 20mins :) <ehsan> jmaher: I haven't really looked at how bad this is on other test suites... <ehsan> but don't have anything against doing it across the board <jmaher> ehsan: I would prefer to do it just for browser-chrome (and devtools would get it as well) <ehsan> sgtm <jmaher> ok, nice <jmaher> armenzg_brb: I don't think this will solve the other issues, it might fix a few cases in devtools e10s, but those are still in cleanup mode outside of the scope of taskcluster/docker <jmaher> ehsan: we have a bunch that are 10 or higher: https://dxr.mozilla.org/mozilla-central/search?q=requestlongertimeout%281&redirect=false&case=false <jmaher> ^ that has some mochitest-plain in there as well, so 12 test cases have a multiplier of 10 <ehsan> jmaher: hmm, ok maybe we can lower those

Is this looking in the right direction? I don't see much improvement (if at all): https://treeherder.mozilla.org/#/jobs?repo=try&group_state=expanded&exclusion_profile=false&revision=986bd96c78cc,e0282328d4cf&filter-searchStr=tc%20mochitest

Comment on attachment 8716408 [details] [diff] [review] Bump mochitest time out from 45 to 90 seconds + reduce browser_chrome multiplies over 5 Review of attachment 8716408 [details] [diff] [review]: ----------------------------------------------------------------- these changes look great, but we are missing the 45 -> 90 second adjustment in this patch.

Not embarassing whatsoever! :) I've got too focused on the multipliers. I will try again later today.

Comment on attachment 8717111 [details] [diff] [review] Bump mochitest time out from 45 to 90 seconds + reduce browser_chrome multiplies over 5 Review of attachment 8717111 [details] [diff] [review]: ----------------------------------------------------------------- Please announce this change on dev-platform and firefox-dev. Thanks!

We've got another change going on which might require this unneeded. We can run the tests on mounted dirs that are run on SSD. As far as I can tell, browser_crashedTabs.js is not timing out anymore [1] For my own record, 60 seconds also works well. [1] https://treeherder.mozilla.org/#/jobs?repo=try&revision=84fbd2ecb0dd&filter-searchStr=browser-chrome&group_state=expanded

I don't think we need this anymore.