Dashboard and alerts¶
Official dashboard¶
The official cnpg grafana dashboard is located here
https://github.com/cloudnative-pg/grafana-dashboards/blob/main/charts/cluster/grafana-dashboard.json
Note: This metric list is based on the current dashboard version. The required metrics may change when the dashboard is updated. Always verify against the latest dashboard version.
Required Metrics¶
To make the official CNPG Grafana dashboard work, you need to ensure the following metrics are available in your Prometheus instance:
CNPG Operator and Cluster Metrics¶
These metrics are provided by CloudNative-PG operator and PostgreSQL clusters:
cnpg_pg_replication_streaming_replicas
cnpg_pg_replication_is_wal_receiver_up
cnpg_pg_replication_lag
cnpg_pg_stat_replication_write_lag_seconds
cnpg_pg_stat_replication_flush_lag_seconds
cnpg_pg_stat_replication_replay_lag_seconds
cnpg_pg_postmaster_start_time
cnpg_pg_stat_database_xact_commit
cnpg_pg_stat_database_xact_rollback
cnpg_backends_total
cnpg_pg_settings_setting
cnpg_pg_replication_in_recovery
cnpg_pg_stat_archiver_seconds_since_last_archival
cnpg_collector_last_available_backup_timestamp
cnpg_collector_postgres_version
cnpg_pg_database_size_bytes
cnpg_collector_first_recoverability_point
Kube State Metrics¶
These metrics are provided by kube-state-metrics:
kube_pod_container_resource_requests
kube_pod_status_ready
kube_pod_container_status_ready
kube_pod_info
kube_node_labels
Kubelet Metrics¶
These metrics are exposed by the Kubelet:
kubelet_volume_stats_available_bytes
kubelet_volume_stats_capacity_bytes
kubelet_volume_stats_inodes_used
kubelet_volume_stats_inodes
cAdvisor Metrics¶
These metrics are provided by cAdvisor (part of Kubelet):
container_memory_working_set_bytes
Controller Runtime Metrics¶
These metrics are provided by the controller-runtime library (used by CNPG operator):
controller_runtime_reconcile_total
Recording Rules¶
The dashboard requires this recording rule to be created:
node_namespace_pod_container:container_cpu_usage_seconds_total:sum_irate
This rule aggregates CPU usage rates by node, namespace, pod, and container dimensions.
Required source metric: The recording rule depends on container_cpu_usage_seconds_total which is provided by cAdvisor (part of Kubelet). This metric must be available in your Prometheus instance for the recording rule to work.
Alerts¶
- Default provided alerts