Details
-
Bug
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
2.4
-
None
-
None
Description
Tracking partition update counters and cache sizes may not properly work if cache is using readThrough behavior.
Read requests to such cache can increment update counters or cache sizes not on all nodes serving such cache in case if data in underlying storage is changed.
It means that update counter or cache size will be incremented only on partition where we followed such request (primary or any random node).
BackupPostProcessingClosure should use preload=false for entry. In other case it can increment update counter for read request while data is not changed.