Monitoring Sidecar

As of v0.12, Backup pods will run with 2 containers by default: the main backup process and a monitoring sidecar container.

The monitoring sidecar is a lightweight process that will periodically check if the main backup process is behaving as expected.

Currently, it checks for the following issues:

• whether the backup process is producing metrics for all defined topics;
• whether the backup process is making progress on all topics;

If one of the tests fails, the container will emit error logs. More checks will be added in upcoming releases.

To perform its job, the monitoring sidecar will create a single additional connection to the cluster using the same credentials as the main backup process. It will periodically consume a single record from the eventhub for all backed-up topics.

Disabling the monitoring sidecar

Disabling the monitoring sidecar can be done by setting spec.monitoring.enabled: false in your Backup definition:

kind: Backup
metadata:
  name: gcs-backup
spec:
  enabled: true
  source: "my-kafka-endpoint"
  sink: "gcs-storage"
  monitoring:
    enabled: false

Setting resource limits on the monitoring sidecar

Resource limits may be enforced the same way as the main backup container under the spec.monitoring.resources property.

For instance:

kind: Backup
metadata:
  name: gcs-backup
spec:
  enabled: true
  source: "my-kafka-endpoint"
  sink: "gcs-storage"
  monitoring:
    # Optional resource requests/limits
    resources:
      limits:
        cpu: 100m
        memory: 500Mi
      requests:
        cpu: 100m
        memory: 100Mi