The Cassandra cluster should be closely monitored using the usual Cassandra JMX beans, e.g.:
- Number of UP/DOWN nodes
- Number of Read/Write requests per second
- Read/Write latency
The Provider exposes metrics on its management port 7979 with path /metrics.
- DropWizard timer for successful access token requests by realm.
- DropWizard timer for failed access token requests by realm and error type.
The Provider is mostly CPU-bound by its BCrypt password checking and JWT signing operations. Monitoring the CPU usage percentage is therefore strongly recommended.
The Token Info exposes metrics on its metrics port 9020 with path /metrics.
- Number of public keys in memory.
- Timer for the JWT handler (one timer per realm).
- Timer for the proxy handler (includes cached results and upstream calls).
- Number of upstream cache hits.
- Number of upstream cache misses.
- Number of upstream cache misses because of expiration.
- Timer for calls to the upstream tokeninfo. Cached responses are not measured here.
The Token Info is mostly CPU-bound by its JWT signature verification operations. Monitoring the CPU usage percentage is therefore strongly recommended.