1602768 - taskclusteretl.timing stopped to get 'superseded' tasks on 2019-10-08

Status: RESOLVED FIXED

(Data Platform and Tools :: General, defect, P1)

Description

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

The taskclusteretl.timing table stopped got get tasks with result superseded on 2019-10-08 while Treeherder's table reports it until yesterday when it got disabled (bug 1602446).

BigQuery: https://sql.telemetry.mozilla.org/queries/66854/source
Latest push: 2019-10-08 10:56

Treeherder: https://sql.telemetry.mozilla.org/queries/66855/source
Latest push: 2019-12-09 16:35

Comment 1

[Mike Trinkala [:trink]]

The way exceptions are handle changed here: https://github.com/mozilla-services/lua_sandbox_extensions/commit/922ed4e2c29a207a19401f58f3d1d97b646d2b17#diff-355470dd5d3a0a0058d77c23eee0471a

Since the exception logs are no longer parsed the code that sets the result is now being skipped defaulting it to 'success'. The tasks are not missing but the derived/normalized result field is mis-classified.

The code will be corrected and the mis-classification in the DB will be corrected.

Comment 2

[Mike Trinkala [:trink]]

Updated, it worked out well since the change in behavior was discovered but the task_resolution table is a better place to check this since it contains the actual reasonResolved and is a much smaller data scan so it costs much less (600MB vs 200GB).

BTW: still seeing 'superseded' entries as of yesterday