Bug suggestions for full-line search appear to only search the first 125 characters of the summary

NEW
Unassigned

Status

Tree Management
Treeherder: Log Parsing & Classification
a year ago
3 months ago

People

(Reporter: philor, Unassigned)

Tracking

Details

(Reporter)

Description

a year ago
There's a whole set of bugs for "Pulling docker image" failures, https://bugzilla.mozilla.org/buglist.cgi?quicksearch=pulling%20docker%20image%20kw%3Aintermittent

Based on (an earlier version of) https://treeherder.mozilla.org/api/project/mozilla-inbound/artifact/?format=api&job_id=35330650&name=Bug+suggestions I finally got them suggested, by including the first two characters of the taskId hash for every current hash (there appear to be two current hashes at a time, but then a commit triggers image builds and there are four until it merges around and settles down, or possibly six while it is merging around), trimming the rest to have enough space to include the actual failure message, which appears inconveniently at the end of the message, in the bug suggestions.

That worked just fine until I put the main current one, QO, at the end of "Uz|EN|Hf|U9|QO". Then nothing was being suggested for failures where the failure message included the QO hash. As an experiment, I moved QO to the front of that list in one bug, and now it is suggested for any QO failure, so apparently only the first 124 or 125 characters of the bug summary are actually being matched for full-line searches. (By the time you read this, you won't be able to see that actual thing failing in production, since I need to have all of them suggested, not just one, so within a few minutes they'll all be switched to have QO first.)
(Reporter)

Comment 1

a year ago
And as usual for my bugs, when I reread that I found it confusing.

If you add to the bug suggestions test the failure line

[taskcluster:error] Pulling docker image {"path":"public/image.tar","type":"task-image","taskId":"QOCI4V_cS3qQgnhg3OJnQw"} has failed: Error loading docker image. HTTP code is 404 which indicates error: undefined - Untar re-exec error: exit status 1: output: no such file or directory

and the bug summary

Intermittent [taskcluster:error] Pulling docker image {"path":"public/image.tar","type":"task-image","taskId":"Uz|EN|Hf|U9|QO"} has failed: HTTP code is 404 which indicates error: undefined - Untar re-exec error: exit status 1: output: no such file or di

then you should find (to the extent the test matches the production behavior) that the bug will not be suggested even though it should be. Move QO| in front of Uz in the summary and take it off the end of that chunk of summary, and you should find that it is suggested.
The best solution here is just to stop including the task ID in the main error line. The error messages shouldn't be unique per instance (the task ID can be listed on another line).
Component: Treeherder → Docker Images
Product: Tree Management → Taskcluster
Summary: Bug suggestions for full-line search appear to only search the first 125 characters of the summary → "Pulling docker image" error messages shouldn't include the TaskID in the main string
Version: --- → unspecified
Blocks: 778688
Component: Docker Images → Worker
Looks like this is possibly a duplicate of bug 1300332 where a patch has already been submitted for review.

Updated

a year ago
See Also: → bug 1300332
Ah great :-)

In which case I'll make this about the originally filed issue again, albeit it's lower priority (and likely just some quirk of mysql full text indexes).
No longer blocks: 778688
Component: Worker → Treeherder: Data Ingestion
Product: Taskcluster → Tree Management
Summary: "Pulling docker image" error messages shouldn't include the TaskID in the main string → Bug suggestions for full-line search appear to only search the first 125 characters of the summary
Version: unspecified → ---
(Reporter)

Comment 5

a year ago
Also fails to work:

search: [taskcluster:error] Task timeout after 5400 seconds. Force killing container.

bug summary that should be found: Intermittent [taskcluster:error] Task timeout after 3600, 5400, 7200, 10800 seconds. Force killing container.
Component: Treeherder: Data Ingestion → Treeherder: Log Parsing & Classification
You need to log in before you can comment on or make changes to this bug.