Follow-up for "Geo scans over 150 000 `geo_event_log` tuples per minute using a sequential scan"
Background
In https://gitlab.com/gitlab-org/gitlab-ee/issues/6621 we noticed a high load of sequential scans on the table geo_event_log
:
This was caused by the Geo::PruneEventLogWorker
. The cause of this is modified in https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/6175.
To avoid this worker would cause more load, it was disabled by taking the Exclusive Lease
Action plan
When an RC of 11.4 (containing https://gitlab.com/gitlab-org/gitlab-ee/merge_requests/6175) is running stable on gitlab.com, release the exclusive lease again, and monitor the load the worker generates.
Since the above mentioned MR also includes a background migration to remove orphaned event log rows, we might want to wait at least until that migration is completed.
Edited by Toon Claes