no linux32 partial gen tasks for beta
Categories
(Release Engineering :: Release Automation, defect)
Tracking
(Root Cause:Infrastructure/Build Error, firefox73blocking verified, firefox74 fixed)
Root Cause | Infrastructure/Build Error |
People
(Reporter: jlund, Assigned: egao)
References
Details
Attachments
(1 file)
we only had a linux64 partial generation task. no linux32 equivalent of: https://firefox-ci-tc.services.mozilla.com/tasks/dCLNypu9RCGxuLw3TglaNw
6:06 PM and for devedition, we did have a partial gen task for linux32: https://firefox-ci-tc.services.mozilla.com/tasks/Yll4fhZ1TPmuL-6yCyjRFg
I do see linux32 in the platform list for partial gen: https://hg.mozilla.org/releases/mozilla-beta/file/tip/taskcluster/ci/partials/kind.yml#l33
this bug was discovered after bouncer check and UV tasks started to fail:
Reporter | ||
Comment 1•5 years ago
|
||
notably, linux32 was demoted to tier 2 but only on central: Bug 1599197 - demote linux32 to tier-2, run builds not on autoland and try by default, only m-c
Reporter | ||
Comment 2•5 years ago
|
||
we can ship without linux32 partials for b1. However, it was pointed out that that may cause errors in our update verify checks for many betas throughout this current cycle
Assignee | ||
Comment 3•5 years ago
|
||
I suspect it might be my changes in https://phabricator.services.mozilla.com/D56605, as one of the first steps to put ubuntu1804-test
docker image into production use?
Reporter | ||
Comment 4•5 years ago
•
|
||
the decision task log is not showing much of a difference when grepping for partial
and diffing beta vs devedition
[task 2020-01-06T20:53:26.885Z] loading kind `partials` from `taskcluster/ci/partials`
[task 2020-01-06T20:53:27.165Z] loading kind `partials-signing` from `taskcluster/ci/partials-signing`
[task 2020-01-06T20:53:27.638Z] No partials defined! Bouncer submission task won't send any partial-related entry for "firefox"
[task 2020-01-06T20:53:27.639Z] No partials defined! Bouncer submission task won't send any partial-related entry for "fennec"
[task 2020-01-06T20:53:27.639Z] No partials defined! Bouncer submission task won't send any partial-related entry for "devedition"
[task 2020-01-06T20:53:33.087Z] Loading tasks for kind partials
[task 2020-01-06T20:53:33.458Z] Generated 1151 tasks for kind partials
[task 2020-01-06T20:53:33.523Z] Loading tasks for kind partials-signing
[task 2020-01-06T20:53:33.797Z] Generated 1151 tasks for kind partials-signing
[task 2020-01-06T20:54:24.641Z] Creating task with taskId HzO9C0wKSZ2sjbL9RWHWYw for partials-win32-shippable/opt
[task 2020-01-06T20:54:24.671Z] Creating task with taskId cyWFA2F6SLao3sHsRbd9EQ for partials-win64-shippable/opt
[task 2020-01-06T20:54:25.690Z] Creating task with taskId NitpFBh2QNi6FieaEGhuOw for partials-macosx64-shippable/opt
[task 2020-01-06T20:54:25.743Z] Creating task with taskId dCLNypu9RCGxuLw3TglaNw for partials-linux64-shippable/opt
[task 2020-01-06T20:54:25.931Z] Creating task with taskId SrvuplmoQZubjNslQ-w30A for partials-win64-aarch64-shippable/opt
[task 2020-01-06T20:54:26.734Z] Creating task with taskId P6xH2-qVQJOefI1orEcf2w for partials-signing-macosx64-shippable/opt
[task 2020-01-06T20:54:27.171Z] Creating task with taskId RPvkI19MTxm7RJv1rO-Aew for partials-signing-win64-shippable/opt
.... etc
[task 2020-01-06T20:53:20.327Z] loading kind `partials` from `taskcluster/ci/partials`
[task 2020-01-06T20:53:20.927Z] loading kind `partials-signing` from `taskcluster/ci/partials-signing`
[task 2020-01-06T20:53:21.310Z] No partials defined! Bouncer submission task won't send any partial-related entry for "firefox"
[task 2020-01-06T20:53:21.311Z] No partials defined! Bouncer submission task won't send any partial-related entry for "fennec"
[task 2020-01-06T20:53:21.311Z] No partials defined! Bouncer submission task won't send any partial-related entry for "devedition"
[task 2020-01-06T20:53:27.122Z] Loading tasks for kind partials
[task 2020-01-06T20:53:27.521Z] Generated 1151 tasks for kind partials
[task 2020-01-06T20:53:27.593Z] Loading tasks for kind partials-signing
[task 2020-01-06T20:53:27.888Z] Generated 1151 tasks for kind partials-signing
[task 2020-01-06T20:54:21.882Z] Creating task with taskId Yll4fhZ1TPmuL-6yCyjRFg for partials-linux-devedition-nightly/opt
[task 2020-01-06T20:54:22.416Z] Creating task with taskId aTeNiSeWQXORuDuEd5DYGQ for partials-macosx64-devedition-nightly/opt
[task 2020-01-06T20:54:22.459Z] Creating task with taskId VQgvWR3aTluY_DwOsM_DXQ for partials-win32-devedition-nightly/opt
[task 2020-01-06T20:54:22.621Z] Creating task with taskId QipWqY-VT5OsfQGpcNd8Dg for partials-win64-aarch64-devedition-nightly/opt
[task 2020-01-06T20:54:23.016Z] Creating task with taskId KPVK2_uPTCa1TCpYbmpvkw for partials-linux64-devedition-nightly/opt
[task 2020-01-06T20:54:23.045Z] Creating task with taskId RoQ_p_7OTdGgP9WlRF7u4A for partials-win64-devedition-nightly/opt
[task 2020-01-06T20:54:23.171Z] Creating task with taskId LLvDdsI-RHufvxcjZNYrAA for partials-signing-linux-devedition-nightly/opt
[task 2020-01-06T20:54:23.488Z] Creating task with taskId JgC7rDrATRClRetcby20TQ for partials-signing-win64-devedition-nightly/opt
[task 2020-01-06T20:54:23.970Z] Creating task with taskId RIkou1pTSSu8kc7XTubKuQ for partials-signing-macosx64-devedition-nightly/opt
.... etc
Assignee | ||
Comment 5•5 years ago
|
||
Updated•5 years ago
|
Comment 7•5 years ago
|
||
bugherder uplift |
Reporter | ||
Comment 8•5 years ago
|
||
backout in comment 6 was needed because it turns out we may actually depend on treeherder names.
https://searchfox.org/mozilla-central/source/taskcluster/taskgraph/transforms/partials.py#71
Which is something we should postmortem on. Ideally we don't depend on Treeherder labels to determine what release tasks we trigger.
Updated•5 years ago
|
Comment 9•5 years ago
|
||
bugherder |
Reporter | ||
Comment 10•5 years ago
|
||
(In reply to Jordan Lund (:jlund) from comment #8)
backout in comment 6 was needed because it turns out we may actually depend on treeherder names.
https://searchfox.org/mozilla-central/source/taskcluster/taskgraph/transforms/partials.py#71
Which is something we should postmortem on. Ideally we don't depend on Treeherder labels to determine what release tasks we trigger.
@sfraser - do you have any context or opinion about depending on treeherder labels for determining which release partial tasks to run?
Comment 11•5 years ago
|
||
(In reply to Jordan Lund (:jlund) from comment #10)
Which is something we should postmortem on. Ideally we don't depend on Treeherder labels to determine what release tasks we trigger.
@sfraser - do you have any context or opinion about depending on treeherder labels for determining which release partial tasks to run?
We don't currently have a nicely parseable target platform in the task metadata, and we probably should. The current tags
.os
is the workerType's operating system, which may cause confusion. I'd be in favour of adding the target platform and other fields to another area, such as metadata, tags or extra. I imagine we'd want to encode 'opt', 'debug', 'asan' and so on in there as a different key, rather than all mashed into one pseudoplatform name.
Comment 12•5 years ago
|
||
We actually do have the build platform in an attribute, and I have a patch switching to use it mostly ready. (There is some dead code in funsize that passes platform around, but doesn't consume it that I need to still adjust).
Comment 13•5 years ago
|
||
(In reply to Tom Prince [:tomprince] from comment #12)
We actually do have the build platform in an attribute, and I have a patch switching to use it mostly ready. (There is some dead code in funsize that passes platform around, but doesn't consume it that I need to still adjust).
There are around ~129 unique values for build_platform, and it encodes information that's a bit too specific for the partial generation's needs, as we need to map it to what Balrog thinks of as a platform, which would mean reliably removing things like '-devedition-nightly', '-shippable', '-ccov' and so on. Removing text based on '-' seems like it would work at first glance, but it also makes me wonder why we're not encoding this in a machine readable way to start with, and building the more complex strings when we need them.
Comment 14•5 years ago
|
||
(In reply to Simon Fraser [:sfraser] ⌚️GMT from comment #13)
There are around ~129 unique values for build_platform
I'm guessing once we strip this down to platforms we actually create partials for (shippable and nightly only, desktop only), that number will be significantly lower.
Updated•5 years ago
|
Comment 15•5 years ago
|
||
Please specify a root cause for this bug. See :tmaity for more information.
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Comment 16•5 years ago
|
||
For this bug the issue was my changes to infrastructure configuration. I selected Testing Error but now I see there is an entry for Infrastructure, so I'd like the classification to be changed.
Comment 17•5 years ago
|
||
(In reply to Edwin Takahashi (:egao) from comment #16)
For this bug the issue was my changes to infrastructure configuration. I selected Testing Error but now I see there is an entry for Infrastructure, so I'd like the classification to be changed.
Hi Edwin,
You should be able to change it now.
Comment 18•5 years ago
|
||
Updated RCA as Infrastructure as per comment#16
Updated•1 month ago
|
Description
•