Deploy PostgreSQL

PostgreSQL is used by Gnocchi to index the data collected and sent by Ceilometer.

Install the Postgres Operator

We are using the Zalando postgres-operator which offers easy to run and highly-available PostgreSQL clusters on Kubernetes.

The following command to install the operator must be run twice, at least for now, due to a race condition with the way kubectl handles the CRD installation.

kubectl kustomize --enable-helm /opt/genestack/base-kustomize/postgres-operator | kubectl apply -f -
sleep 10
kubectl kustomize --enable-helm /opt/genestack/base-kustomize/postgres-operator | kubectl apply -f -

Create the PostgreSQL Cluster

With kubectl (Recommended)With kubectl kustomize Overlay

Customize as needed

Be sure to modify the cluster parameters to suit your needs. The below values should work fine for a small lab or staging envionrment, however more disk space and other changes may be required in production.

kubectl apply -f - <<EOF
apiVersion: "acid.zalan.do/v1"
kind: postgresql
metadata:
  name: postgres-cluster
  namespace: openstack
spec:
  dockerImage: ghcr.io/zalando/spilo-16:3.2-p3
  teamId: "acid"
  numberOfInstances: 3
  postgresql:
    version: "16"
    parameters:
      shared_buffers: "2GB"
      max_connections: "1024"
      log_statement: "all"
  volume:
    size: 40Gi
  nodeAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      nodeSelectorTerms:
        - matchExpressions:
          - key: node-role.kubernetes.io/worker
            operator: In
            values:
            - worker
EOF

Two overlays exist - base which includes 3 replicas, and an aio overlay which has a single replica and less default resource utilization.

kubectl kustomize /opt/genestack/base-kustomize/postgres-cluster/base | kubectl apply -f -