Details
-
Suggestion
-
Resolution: Duplicate
-
None
-
None
Description
When FishEye detects number of repositories that need to be indexed (number of repositories added around the same time, or instance being migrated to another box with no FishEye caches), each Initial Pinger thread may spend a long time (hours, days) indexing particular repository. That leads to starvation of other repositories.
Although in the short term it seems sensible to keep Initial Pinger threads "sticky", to avoid context switching, perhaps FishEye could ignore this "stickiness" after some threshold (e.g. 1h) so other repositories would have a chance to get indexed too.
Attachments
Issue Links
- relates to
-
FE-5472 Setting a short poll period, and having repositories that take a long time to check for changes, might cause other repositories never to be scheduled
- Closed