Monitor an instance
The Operator can inject a redis_exporter sidecar. The exporter listens on port
9121 and serves metrics at /metrics.
Configure the exporter
The high-level API enables the exporter unless spec.exporter.disable is true.
Set resources explicitly for predictable scheduling:
Disable it only when another monitoring mechanism is installed:
Verify exporter health
For Cluster architecture, the instance Service exposes the metrics port:
In another terminal:
For Failover or Replica, discover the headless metrics Service, whose name starts
with rfr-, and port-forward it:
The Failover metrics Service exists even when the exporter is disabled, because
it is also the headless governing Service of the data StatefulSet. Before
treating the Service as a scrape target, verify that the selected Pods actually
contain a ready exporter container and that /metrics responds.
You can also inspect the sidecar directly:
Integrate scraping
The ServiceMonitor shipped with the Operator selects Operator metrics. It does
not by itself scrape Valkey data-plane Services. Create a ServiceMonitor,
PodMonitor, or equivalent scrape configuration that matches the Services in
your environment, and verify target health in your monitoring system.
When TLS and access control list (ACL) authentication are enabled, the Operator
supplies the exporter with its ACL identity and TLS files, and the exporter
resolves its password from the referenced Kubernetes Secret through the API
rather than receiving it in an environment variable. Do not copy credentials
into scrape configuration.
The implementation uses redis_exporter; validate every required metric against
the delivered exporter image and the selected Valkey line, especially 9.1.