StorageGRID high CPU utilization by Cassandra database result in increased S3 client latency
Applies to
- NetApp StorageGRID
- Software release 11.6 and later versions
Issue
- All the metrics presented below can be found in StorageGRID Grid Manager under Support > Metrics.
- Storage node shows Cassandra CPU utilization greater than 85% with minimal I/O wait. This can be found under Node (Internal Use) Grafana dashboard.
- Cassandra ReadStage is consistently at its maximum level. This can be found under Cassandra Node Overview Grafana dashboard.

- A single Cassandra table shows extremely high latency in the hundreds of seconds. This can be found under Cassandra Node Overview Grafana dashboard.
- Below example shows object_by_version with such high latency. Note that it could be any table.
- In the node's Cassandra directory, the following command returns more than 100 entries. Consider checking the rotated log file (i.e
jstack.log.1
).
# cd /var/local/log/cassandra/jstack/
# grep -c Murmur3 jstack.log
8393