Details
-
Bug
-
Resolution: Fixed
-
Critical
-
Cheshire-Cat
-
Untriaged
-
1
-
Unknown
-
CX Sprint 237
Description
As observed in MB-44050, at the start of rebalance, we attempt to gather the aggregated storage state of all datasets in the cluster. This times out after 10 minutes for 999 datasets each with a (i.e. 999) secondary index.
- There's no need to do this all at once for all 999 datasets, we should do it per-dataset, or in reasonable batches
- likely we should do this in the background as we otherwise progress on the rebalance-- i.e. don't attempt to know the full sizes up-front- instead calculate them incrementally, allowing the dataset progress to become more accurate as we obtain more answers
Attachments
Issue Links
- is triggered by
-
MB-44050 Analytics rebalance tests with 999 collections failed on build 7.0.0-4342
- Closed