Follow-up from "Remove :failed category when receiving a non-success webhook log"

About

This issue was created from this discussion !166329 (comment 2402019512):

We set the backoff_count to be 3 less than the recent_failures because we only increment backoff_count after we reach 3 recent_failures.

You can see this relationship between the two numbers in the logs https://log.gprd.gitlab.net/app/r/s/7LCpF.

As a follow-up I think we can remove backoff_count and just refer to recent_failures (minus FAILURE_THRESHOLD from it where we would normally refer to backoff_count).

Proposal

The first MR !187077 (merged) to remove backoff_count was merged in 17.11.

The next step will be 18.2 - the next required stop, as we will need to finalize the MigrateOldDisabledWebHookToNewState background migration that is setting backoff_count for records.

The remaining work is in 18.2 or later:

• Finalize MigrateOldDisabledWebHookToNewState https://docs.gitlab.com/development/database/batched_background_migrations. This will ensure that our migration to drop the backoff_count will not interfere with any remaining batches from that migration.
• Drop backoff_count https://docs.gitlab.com/development/database/avoiding_downtime_in_migrations/#dropping-the-column-release-m1

In the milestone after we drop backoff_count: Remove the ignore rule https://docs.gitlab.com/development/database/avoiding_downtime_in_migrations/#removing-the-ignore-rule-release-m2