Skip to content

README.md

Latest commit

 

History

History
86 lines (57 loc) · 5.17 KB

README.md

File metadata and controls

86 lines (57 loc) · 5.17 KB

Deploy — EDB Postgres on OpenShift (CloudNativePG)

YAML layout to install the EnterpriseDB operator distribution and a small sample Cluster.

On Red Hat OpenShift, prefer Operator Lifecycle Manager (OLM) via db-deploy/olm-openshift/ (OperatorHub subscription flow). Use the manifest bundle under db-deploy/operator/ for a pinned YAML install without OperatorHub when you need non-OLM deployment.

Layout

Path Purpose
olm-openshift/ Preferred on OpenShift: OLM Subscription (cluster-wide) and an example OperatorGroup + Subscription for scoped installs — EDB OpenShift / oc CLI. See olm-openshift/README.md.
operator/kustomization.yaml Non-OLM manifest install: pinned operator manifest from get.enterprisedb.io (creates postgresql-operator-system and CRDs).
sample-cluster/ Example namespace, app credentials secret, and Cluster CR (edb-postgres / postgresql).
cross-cluster/ Passive replica (streaming) across two clusters: Route on primary + replica Cluster + script — see cross-cluster/README.md.

Prerequisites

  • kubectl / oc configured for the target cluster (cluster-admin or equivalent to install operators and CRDs).
  • OpenShift (OLM): EDB pull secret in openshift-operators and subscription steps — olm-openshift/README.md.
  • OpenShift (manifest install from operator/ only): if the controller pod fails on SCC, grant a suitable SCC to postgresql-operator-manager in postgresql-operator-system and restart the deployment — docs/openshift-edb-operator-smoke-test.md.

Install operator

OpenShift — OLM (recommended)

oc apply -k db-deploy/olm-openshift

Verify and complete pull-secret / CSV approval steps in olm-openshift/README.md. For multi-namespace or single-namespace operator placement, use olm-openshift/operatorgroup-multinamespace.example.yaml instead of the kustomize overlay.

OpenShift / non-OLM — manifest bundle

Use server-side apply so large CRDs apply cleanly:

kubectl apply --server-side --force-conflicts -k db-deploy/operator
kubectl rollout status deployment/postgresql-operator-controller-manager \
  -n postgresql-operator-system --timeout=300s

Edit sample workload

  1. Set a strong password in sample-cluster/base/app-db-credentials.secret.yaml.
  2. Set spec.storage.storageClass in sample-cluster/base/cluster.yaml if your cluster has no default StorageClass.

For EDB registry images, create edb-pull-secret (or your name) and either edit base/cluster.yaml to match base/cluster-edb-registry.yaml, or swap the filename in sample-cluster/base/kustomization.yaml resources.

Apply sample cluster

Apply after the operator install is healthy (Cluster CRD available). The sample postgresql cluster uses spec.instances: 2 (primary + one hot standby in the same OpenShift cluster).

Base (omit storageClass and rely on the cluster default StorageClass):

kubectl apply -k db-deploy/sample-cluster
kubectl get cluster,pods -n edb-postgres -w

Overlays (set spec.storage.storageClass for clusters without a suitable default). Kustomize’s default load restrictor blocks kubectl apply -k on these paths because they reference ../../base. Build with LoadRestrictionsNone, then apply:

kubectl kustomize --load-restrictor LoadRestrictionsNone db-deploy/sample-cluster/overlays/overlay-lvms-vg1 | kubectl apply -f -
# or CRC-style default SC:
kubectl kustomize --load-restrictor LoadRestrictionsNone db-deploy/sample-cluster/overlays/overlay-topolvm-provisioner | kubectl apply -f -

OpenShift’s oc kustomize supports the same flag, for example:
oc kustomize --load-restrictor LoadRestrictionsNone db-deploy/sample-cluster/overlays/overlay-lvms-vg1 | oc apply -f -.

Path StorageClass
db-deploy/sample-cluster/overlays/overlay-lvms-vg1 lvms-vg1
db-deploy/sample-cluster/overlays/overlay-topolvm-provisioner topolvm-provisioner

Repeat OLM install + sample overlay on each OpenShift if you want two independent two-node Postgres clusters — use explicit --kubeconfig / --context per cluster. See olm-openshift/README.md.

Cross-cluster passive replica

Use cross-cluster/README.md for prerequisites, anonymity notes, and the full flow. Short version:

  1. Install the operator on both clusters; create a healthy primary Cluster (e.g. db-deploy/sample-cluster/ overlays).
  2. Export PRIMARY_CONTEXT and REPLICA_CONTEXT (and optional split kubeconfigs).
  3. Run db-deploy/cross-cluster/scripts/sync-passive-replica.sh.

Official reference: EDB — Replica clusters and EDB Postgres on OpenShift (operator docs).