1230763 - Move Thunderbird OS X tests from "yosemite r5" to "yosemite r7"

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: CIDuty, task)

Description

[aleth [:aleth]]

The r5 build slaves are being decommissioned.

Comment 1

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763.patch

Attached you can find the patch. Removed yosemite and added yosemite_r7

Comment 2

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.patch

Attached the correct patch

Comment 3

[Kim Moir [:kmoir] ET]

Comment on attachment 8697979 [details] [diff] [review]
bug1230763_thunderbird_config.py.patch

I would recommend asking the person who opened this bug if they want us to move all jobs immediately or if we want this to ride the trains.  If we want to move all jobs, we can remove the old yosemite definitions. Also, please provide a builder diff :-)

Comment 4

[Vlad Ciobancai [:vladC] [:vciobancai]]

Do you want to move all the jobs immediately from yosemite r5 platform to yosemite r7 platform or do you want to run in parallel for both platforms ?

Comment 5

[aleth [:aleth]]

Forwarding to Callek.

Comment 6

[Justin Wood (:Callek)]

(In reply to aleth [:aleth] from comment #5)
> Forwarding to Callek.

from a moco releng standpoint, we'll need to drop off all yosemite r5's soon. But I don't know if the Thunderbird team considers it a quick drop-in replacement, or if they need any parallelization.

I'll poke :rkent for the final say from the Thunderbird Teams position

Comment 7

[Kent James (:rkent)]

Wow, I don't have any idea what this is about. :aleth has been on top of build issues for us, so at the moment I'd either defer to him, or ask :fallen if he knows anything about this.

I can research an opinion if needed, but I would trust :aleth to determine if we need to collect more opinions on this.

Comment 8

[aleth [:aleth]]

I assumed it was a drop-in replacement. Does anything specifically depend on the builds happening on r5 as opposed to r7 (ie. does the build config differ)? In particular, do ESR 38 builds have to happen on the older r5 machines for some reason?

Comment 9

[Phil Ringnalda (:philor)]

It's not builds (they actually happen on machines running 10.7), it's tests, and while it's less likely that Tb will have failures due to the change than it was that Firefox would, it's still possible. Firefox had to land patches to get tests to pass on 10.10.5, and thus is letting those patches merge down toward mozilla-release and switching as it goes, and thus keeping esr38 on 10.10.2, but because the 10.10.5 machines are physically being put into the same space where the 10.10.2 machines were, they want to get as much as possible off 10.10.2 to be able to get down to the absolute minimum possible number of slaves that haven't been swapped out.

So the question is: "do you think your tests on 10.10 will be fine, and if you don't think they will be will you be willing to land changes to them or disable them on release branches?"

Comment 10

[aleth [:aleth]]

(In reply to Phil Ringnalda (:philor) from comment #9)
> So the question is: "do you think your tests on 10.10 will be fine, and if
> you don't think they will be will you be willing to land changes to them or
> disable them on release branches?"

Thanks for the explanation!

If possible, I think it would be best if c-* matched the corresponding m-* in this (in particular for esr38), especially since comm builds also run a lot of gecko tests beside the comm-specific ones, and this would minimize failures solely due to differing sdk versions between TB and Firefox builds.

Comment 11

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.patch

Updated the patch

Comment 12

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.output

Attached the output

Comment 13

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.patch

Attached the correct patch

Comment 14

[Kim Moir [:kmoir] ET]

Comment on attachment 8698929 [details] [diff] [review]
bug1230763_thunderbird_config.py.patch

I think we need to modify this patch so that the it matches the requirements in comment #10 - that the trunk jobs run on 10.10.5 and non-trunk jobs run on 10.10. So the 10.10 jobs for comm-central and try-comm-central should be removed, and the jobs for  comm-esr38, comm-aurora and comm-beta remain on 10.10.

Comment 15

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.output

Comment 16

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.patch

Updated the patch

Comment 17

[Kim Moir [:kmoir] ET]

Comment on attachment 8698988 [details] [diff] [review]
bug1230763_thunderbird_config.py.patch

You can see the current version of gecko here

http://whattrainisitnow.com/

Not sure why you are looking at version 38

why don't you create two variables to list the branches you want to enable like we did in the firefox configs similar to this

http://hg.mozilla.org/build/buildbot-configs/file/tip/mozilla-tests/config.py#l2601

(note that we have had a merge since then so the version specified should be different)

We don't want to have an explicit list like this
 branch_exclusions=['comm-esr38', 'comm-aurora', 'comm-beta'])
but rather construct a list based on the gecko version 

In the builder diff, we shouldn't have both 10.10 and 10.10.5 builders on the same branch.  One or the other.

Comment 18

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.patch

updated the patch

Comment 19

[Vlad Ciobancai [:vladC] [:vciobancai]]

Attachment: bug1230763_thunderbird_config.py.output

Attached the output

Comment 20

[Release Engineering SlaveAPI Service]

In production: https://hg.mozilla.org/build/buildbot-configs/rev/37ca97437c03