Monitoring v2.0.0

Suggest edits

Monitoring Instances

The PGD4K operator leverages the monitoring capabilities of the PG4K operator to provide monitoring for PGD instances. Therefore, all the monitoring configuration options available for PG4K are also available for PGD instances.

The PGD4K operator offers a central location for those monitoring settings in the spec.cnp.monitoring section, which will be applied to all nodes in the PGD group.

By default, the PG4K operator provides an HTTP/HTTPS metrics exporter on port 9187 for Prometheus, and both PG4K and PGD4K come with predefined sets of metrics as well as a highly configurable and customizable system to define additional queries via one or more ConfigMap or Secret resources.

Note

Refer to Exposed ports for information about the ports exposed by PGD nodes for monitoring purposes and the related security implications.

Note

Refer to How to inspect the exported metrics for instructions on how to access and inspect the metrics exported by PGD nodes and the operator pod.

Monitoring with the Prometheus operator

Like monitoring PG4K instances, you can monitor PGD nodes using the Prometheus Operator's PodMonitor resource. The recommended approach is to manually create and manage a PodMonitor for each PG4K Cluster. This method provides you with full control over the monitoring configuration and lifecycle.

Refer to monitoring with the Prometheus operator for more details about monitoring PG4K instances with the Prometheus operator, which is also applicable to PGD nodes.

Predefined set of metrics

The PGD4K operator installation manifests come with a predefined ConfigMap, called pgd-operator-default-monitoring, to be used by all Clusters. MONITORING_QUERIES_CONFIGMAP is by default set to pgd-operator-default-monitoring in the operator configuration. This ConfigMap contains a predefined set of BDR-related monitoring queries that are automatically injected into each PGD node's monitoring configuration. These metrics use the cnp_bdr_ prefix. You can customize this ConfigMap to add or modify the default monitoring queries for all PGD nodes.

Note

Refer to predefined set of metrics for more information about the predefined metrics provided by the PG4K operator, which are also applicable to PGD instances.

User-defined metrics

Default monitoring queries work alongside any user-defined custom queries configured in spec.cnp.monitoring. If a user-defined query has the same metric name as a default one, the user-defined query takes precedence.

Here is an example of how to configure the user-defined queries in the PGDGroup manifest. The customized queries exist in the example-monitoring-configmap ConfigMap and example-monitoring-secret Secret, and they will be injected together with the default queries:

apiVersion: pgd.k8s.enterprisedb.io/v1beta1
kind: PGDGroup
metadata:
  name: group-example-with-monitoring
spec:
  [...]
  cnp:
    [...]
    monitoring:
      disableDefaultQueries: false
      customQueriesConfigMap:
        - name: example-monitoring-configmap
          key: custom-queries
      customQueriesSecret:
        - name: example-monitoring-secret
          key: custom-queries
Note

Refer to user-defined metrics for more information about the user-defined metrics provided by the PG4K operator, which are also applicable to PGD instances.

Default Monitoring Queries

The PGD4K operator also supports distributing a set of default monitoring queries to all PGD nodes. This allows you to define a common set of monitoring queries that will be automatically injected into each node's monitoring configuration, ensuring consistent monitoring across the entire PGD group. You can configure the default monitoring queries by providing a ConfigMap or Secret in the operator's namespace, as described in the following sections.

How it works

The operator automatically replicates the source ConfigMap and/or Secret to each PGDGroup namespace as pgd-operator-default-monitoring, and injects them into every node's monitoring configuration. Changes to the source are synced on the next reconciliation cycle.

Configuring default monitoring queries

To configure default monitoring queries, set the following environment variables in the operator configuration:

Environment VariableDescription
MONITORING_QUERIES_CONFIGMAPName of the ConfigMap in the operator namespace containing the default monitoring queries
MONITORING_QUERIES_SECRETName of the Secret in the operator namespace containing the default monitoring queries

Both the ConfigMap and Secret must store the monitoring queries under the queries data key. You can check the default queries by inspecting the pgd-operator-default-monitoring ConfigMap in the operator namespace:

kubectl get configmap pgd-operator-default-monitoring \
    -n pgd-operator-system -o yaml
Note

In the OpenShift environment, the namespace is replaced with openshift-operators instead of pgd-operator-system.

Important

Changes to the operator configuration require an operator restart to take effect. However, once the operator is configured, any updates to the source ConfigMap or Secret content are automatically propagated to all PGDGroup namespaces without restarting.

Disabling default monitoring queries

You can disable default monitoring queries for a specific PGDGroup by setting disableDefaultQueries to true in the spec.cnp.monitoring section:

apiVersion: pgd.k8s.enterprisedb.io/v1beta1
kind: PGDGroup
metadata:
  name: pgdgroup-example
spec:
  [...]
  cnp:
    monitoring:
      disableDefaultQueries: true
  [...]

When default queries are disabled for a PGDGroup, the operator skips creating the default monitoring resources in that namespace and does not inject them into nodes.

Important

The disableDefaultQueries field controls the default metrics from both the PGD4K operator and the PG4K operator. Setting it to true disables the default set of metrics from PG4K as well as the PGD4K operator's default monitoring queries. If you only want to disable the PGD4K operator's default monitoring queries while keeping the PG4K defaults, leave disableDefaultQueries as false (or unset) and instead set the MONITORING_QUERIES_CONFIGMAP and MONITORING_QUERIES_SECRET environment variables to empty values in the operator configuration.

Note

If all PGDGroups in a namespace have default queries disabled (or are deleted), the operator automatically cleans up the pgd-operator-default-monitoring ConfigMap and Secret from that namespace.

Monitoring PGD4K Operator instance

Like the PG4K operator, the PGD operator also internally exposes Prometheus metrics via HTTP on port 8080, named metrics.

Currently, the PGD operator exposes default Kubebuilder metrics. See kubebuilder documentation for more details.

Monitoring on OpenShift

Refer to the Monitoring on OpenShift section for details.

← Prev

PGD Node management

↑ Up

EDB Postgres® AI for CloudNativePG™ Global Cluster

Next →

PGDGroup parting