1097328 - job_guid for coalesced jobs isn't generated correctly

Status: RESOLVED DUPLICATE of bug 1093743

(Tree Management :: Treeherder: Data Ingestion, defect, P1)

Description

[Cameron Dawson [:camd]]

We use the first element in the request_ids to generate the job_guid for running jobs, but we use a comma-separated concatenated string of all the request_ids for complete jobs.  Whichever is the right one, these should probably match.

See line 455 of buildapi.py

Comment 1

[Cameron Dawson [:camd]]

Here's commit for a potential fix: https://github.com/mozilla/treeherder-service/compare/coalesced-job-guid

Mauro: maybe we could chat about whether this was intentional, or if it should match this line:
https://github.com/mozilla/treeherder-service/blob/6b7c16a6ffc4bce44b85d33aabe7115a22d8767b/treeherder/etl/buildapi.py#L65

Comment 2

[Ed Morley [:emorley]]

Could this be the cause of bug 1097310?

I would think that the correct method would be to use just the last request_id for completed jobs (similar to bug 1093743), since if there was another (later) job, it would have to have had a newer request, so wouldn't result in a job guid collision.

Comment 3

[Ed Morley [:emorley]]

Otherwise pending jobs (which will only have a single request_id) will get a different job guid to their later running/completed job, if coalescing occurred, surely?

Comment 4

[Mauro Doglio [:mdoglio]]

:camd I think :edmorley is right, we should probably just use the last element from request_ids and the corresponding value in request_time_list to create the job_guid.

Comment 5

[Cameron Dawson [:camd]]

Taking this because the fix for this is also in https://github.com/mozilla/treeherder-service/pull/359

Comment 6

[Ed Morley [:emorley]]

Bug 1093743 should fix this :-)