1625033 - Create "built from" url artifact to replace TinderboxPrint log line

Status: RESOLVED FIXED

(Thunderbird :: Build Config, task)

Description

[Sarah Clements [:sclements] | PTO until March 30th]

I'm working on a task to discontinue the storage of job artifacts in the Treeherder database. I'd like to switch to retrieving artifacts in the UI directly from a taskcluster API but we're currently only able to retrieve "artifact uploaded" artifacts. A blocker to fully implementing this is TinderboxPrint. Treeherder currently parses[2] raw text logs for TinderboxPrint lines and stores them as artifacts for use in treeherder's job details panel.

[vcs 2020-03-25T18:26:51.244Z] TinderboxPrint:<a href=https://hg.mozilla.org/mozilla-central/rev/c5112a7573ac8a9d388e6253e3305061654b123e title='Built from mozilla-central revision c5112a7573ac8a9d388e6253e3305061654b123e'>c5112a7573ac8a9d388e6253e3305061654b123e</a>

[task 2020-03-25T08:27:34.766Z] 08:27:34     INFO - TinderboxPrint: xpcshell-xpcshell<br/>852/<em class="testfail">1</em>/4
[task 2020-03-25T08:27:34.850Z] 08:27:34     INFO - TinderboxPrint: CPU usage<br/>42.0%
[task 2020-03-25T08:27:34.851Z] 08:27:34     INFO - TinderboxPrint: I/O read bytes / time<br/>133,754,880 / 3,204
[task 2020-03-25T08:27:34.851Z] 08:27:34     INFO - TinderboxPrint: I/O write bytes / time<br/>3,524,530,176 / 514,556
[task 2020-03-25T08:27:34.852Z] 08:27:34     INFO - TinderboxPrint: CPU idle<br/>921.4 (57.6%)
[task 2020-03-25T08:27:34.852Z] 08:27:34     INFO - TinderboxPrint: CPU system<br/>132.4 (8.3%)
[task 2020-03-25T08:27:34.853Z] 08:27:34     INFO - TinderboxPrint: CPU user<br/>539.1 (33.7%)
[task 2020-03-25T08:27:34.853Z] 08:27:34     INFO - TinderboxPrint: Swap in / out<br/>0 / 0

We format them in our API like so:

 {

      "title": "Built from mozilla-central revision c5112a7573ac8a9d388e6253e3305061654b123e",
       "value": "c5112a7573ac8a9d388e6253e3305061654b123e",
       "url": "https://hg.mozilla.org/mozilla-central/rev/c5112a7573ac8a9d388e6253e3305061654b123e "
 }

{
     "title": "Swap in / out",
      "value": "0 / 0",
      "url": null
},

Tom.Prince suggested a structured artifact could be created that supplies that information (ideally for both the "Built by" link and the log lines).

[1] https://searchfox.org/mozilla-central/rev/4d9cd186767978a99dafe77eb536a9525980e118/taskcluster/scripts/run-task#490

[2] https://github.com/mozilla/treeherder/blob/master/treeherder/log_parser/parsers.py#L258

Hopefully some of the existing code in the two links provided could be reused.

Comment 1

[Sarah Clements [:sclements] | PTO until March 30th]

Also up for discussion is whether all of that is still needed or should still be supported, since it seems TinderboxPrint is an old/legacy reference.

Comment 2

[Sarah Clements [:sclements] | PTO until March 30th]

Ed Morley brought up some really good points about TinderboxPrint 2 years ago per: https://bugzilla.mozilla.org/show_bug.cgi?id=1410737#c7

Comment 3

[Sarah Clements [:sclements] | PTO until March 30th]

Hi Rob, Tom.Prince suggested you are someone who could help with this issue. Specifically creating a structured artifact for the "built from ..." urls that Treeherder is parsing from TinderboxPrint log lines?

Comment 4

[Tom Prince [:tomprince]]

I'll expand a bit on my suggestion. I'm not necessarily suggesting having a tinderbox print artifact, but that the data that was using that be put into a structured artifact (i.e. json), that treeherder could present, rather than parsing all the logs.

Comment 5

[Rob Lemley [:rjl]]

Attachment: image.png

I know one of our developers uses those links enough that he noticed when they broke recently. Probably whatever generates the target.txt artifact can produce a JSON file as well. I don't see the lack of pinning on comm-central being resolved anytime soon, though it's been on my mind.

I'll see what I can come up with.

Comment 6

[Sarah Clements [:sclements] | PTO until March 30th]

Rob, I'll be merging the pr on Thursday that will stop storing that link in our database. Do need more time put a patch together?

Comment 7

[Rob Lemley [:rjl]]

Attachment: Create build artifact to generate "Built from" links in Treeherder

Treeherder is getting an update later this week that affects the "Built from" links that
appear in the Job Details panel. Rather than parsing log files, a build artifact will be
used to display these links.
The artifact is a JSON file in the suggested format.

Comment 8

[Rob Lemley [:rjl]]

Attachment: built_from.json sample file

The artifact file name is "built_from.json". Sarah, will this work?

I only did this for Thunderbird, I can port for Firefox as well if needed.

Comment 9

[Sarah Clements [:sclements] | PTO until March 30th]

That works! So from the taskcluster API, it'll look like this?

{ "storageType": "s3", "name": "public/built_from.json", "expires": <date>, "contentType": "application/json" }

Comment 10

[Sarah Clements [:sclements] | PTO until March 30th]

Comment on attachment 9144209 [details]
built_from.json sample file

Perfect :)

Comment 11

[Rob Lemley [:rjl]]

Not sure where 'build' comes from? But it's present in all of the artifacts it looks like.
Here's the try job I ran if it's helpful.
https://treeherder.mozilla.org/#/jobs?repo=try-comm-central&revision=8b06fe04af5e470a536432f5d6f654e2c2d95beb

{
  "storageType": "s3",
  "name": "public/build/built_from.json",
  "expires": "2020-05-27T00:51:52.797Z",
  "contentType": "application/json"
},

(In reply to Sarah Clements [:sclements] from comment #9)

That works! So from the taskcluster API, it'll look like this?

{ "storageType": "s3", "name": "public/built_from.json", "expires": <date>, "contentType": "application/json" }

Comment 12

[Geoff Lankow (:darktrojan) [OOO, back 7 Apr]]

Comment on attachment 9144206 [details] [diff] [review]
Create build artifact to generate "Built from" links in Treeherder

LGTM!

Comment 13

[Sarah Clements [:sclements] | PTO until March 30th]

Rob, can you let me know when this is change is live (a sample task where it's showing up would be helpful). I want to test the changes I made before I merge it.

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/dce2046514b0
Create build artifact to generate "Built from" links in Treeherder. r=darktrojan CLOSED TREE

Comment 15

[Rob Lemley [:rjl]]

This build will have the artifact.
https://treeherder.mozilla.org/#/jobs?repo=comm-central&selectedTaskRun=QHNQh09ZQha4Hqa5dz-AJg-0&revision=dce2046514b0a603c84bbf6cbf1cbc8730faf35e

It just occurred to me, I only have this artifact on build jobs. It would be useful on the Decision task as well, I'll do that in a follow-up somehow.

Comment 16

[Tom Prince [:tomprince]]

I think it would make sense for run-task to be changed to emit this, rather than having it as part of the build process.

Comment 17

[Rob Lemley [:rjl]]

Currently run-task doesn't seem to write any artifacts, and it looks like it doesn't really know where the right place to do so would be with the information it has. Is there a predictable way of finding the right path for artifacts?

UPLOAD_DIR is not always set when run-task executes. Can I rely on using /builds/worker/artifacts? (%TASKDIR%\builds\public for generic-workers?)

(In reply to Tom Prince [:tomprince] from comment #16)

I think it would make sense for run-task to be changed to emit this, rather than having it as part of the build process.

Comment 18

[Tom Prince [:tomprince]]

(In reply to Rob Lemley [:rjl] from comment #17)

Is there a predictable way of finding the right path for artifacts?

The predictable way is to add an argument to run-task to say where to put it, and adjust the taskgraph.transforms.job.run_task transform to pass that argument, and add an additional artifact specification to cover the path it uses; (my initial reaction is that run-task should use a different artifact namespace than "user level" things.

Comment 19

[Rob Lemley [:rjl]]

Attachment: Bug 1625033 - Write out built_from.json in the decision task. r=#thunderbird-build-system-reviewers

Comment 20

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/comm-central/rev/6d2883f4b474
Write out built_from.json in the decision task. r=dandarnell