How to check high java heap usage caused by merkle tree memory leak in StorageGRID 11.4
Applies to
NetApp StorageGRID 11.4
Description
Under certain circumstances, the version of Cassandra contained within the StorageGRID 11.4 codeline (DSE 5.1.14) can exhibit high java heap usage due to a merkle tree memory leak caused by long-finished repairs.
StorageGRID will report:
- High Java Heap alarms/alerts
- Cassandra repair errors in logs, example:
[ERROR] org.apache.cassandra.repair.Validator:268 - Failed creating a merkle tree for [repair #e9a52ab0-9329-11eb-83c1-8bc0b64bffdd on storagegrid