We've been using action.mindy() in a few places when we need a root user, which creates a new admin user ("mindy") every time. When these user creation requests happen at the same time we occasionally get random failures of our e2e test suite due to database deadlocks.
A simple workaround (hopefully) is to use action.root() instead which uses the pre-existing (per api config) root user instead of creating a new mindy admin user.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Open | None | T371913 [EPIC] Track common CI failures | |||
| Resolved | Jakob_WMDE | T372906 Avoid user table deadlocks in e2e tests by using a pre-existing admin user |
Change #1064053 had a related patch set uploaded (by Jakob; author: Jakob):
[mediawiki/extensions/Wikibase@master] REST: Use singleton root user in e2e tests
Change #1064053 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] REST: Use existing root user in e2e tests
Change #1064352 had a related patch set uploaded (by Jakob; author: Jakob):
[mediawiki/extensions/Wikibase@master] REST: Create dedicated AuthTest user before running e2e tests
Change #1064352 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] REST: Create dedicated AuthTest user before running e2e tests