Migrate MongoDBSearch from Public Preview to GA

Before You Begin

GA enforces a minimum supported mongot version, 1.70.1. If your MongoDBSearch resource pins an explicit spec.version earlier than 1.70.1, reconciliation fails after the upgrade with an unsupported-version error. If you don't set spec.version, the Kubernetes Operator supplies its own default (1.70.1 as of this GA release) and you don't need to take any action.

Before you upgrade, check which version, if any, you've pinned:

kubectl get mongodbsearch <name> -n <namespace> -o jsonpath='{.spec.version}'

If the command prints a version earlier than 1.70.1, update spec.version to 1.70.1 or later as part of the manifest changes in the next section.

Update Your MongoDBSearch Configuration

A Public Preview MongoDBSearch custom resource never had a spec.clusters field. At GA, spec.clusters is required, and several fields that used to live at the top level of spec now live only inside spec.clusters[0]. Every existing manifest needs the same mechanical edit, once, regardless of topology.

The examples in this section use a MongoDBSearch resource named my-search in the mongodb namespace. Substitute your own resource's name and namespace.

apiVersion: mongodb.com/v1
kind: MongoDBSearch
metadata:
  name: my-search
  namespace: mongodb
spec:
  source:
    mongodbResourceRef:
      name: my-replica-set
  replicas: 3
  persistence:
    single:
      storage: 20Gi
  resourceRequirements:
    requests:
      cpu: "1"
      memory: 2Gi
    limits:
      cpu: "2"
      memory: 4Gi
  statefulSet:
    spec:
      template:
        spec:
          nodeSelector:
            disktype: ssd
  loadBalancer:
    managed:
      replicas: 2
  jvmFlags:
    - "-Xmx2g"

apiVersion: mongodb.com/v1
kind: MongoDBSearch
metadata:
  name: my-search
  namespace: mongodb
spec:
  source:
    mongodbResourceRef:
      name: my-replica-set
  clusters:
    - replicas: 3
      persistence:
        single:
          storage: 20Gi
      resourceRequirements:
        requests:
          cpu: "1"
          memory: 2Gi
        limits:
          cpu: "2"
          memory: 4Gi
      statefulSet:
        spec:
          template:
            spec:
              nodeSelector:
                disktype: ssd
      loadBalancer:
        managed:
          replicas: 2
      jvmFlags:
        - "-Xmx2g"

spec:
  prometheus:
    port: 9946

spec:
  observability:
    prometheus:
      port: 9946

spec:
  observability:
    prometheus:
      mode: disabled

apiVersion: v1
kind: Secret
metadata:
  name: my-search-external-ca
  namespace: mongodb
type: Opaque
stringData:
  ca.crt: |
    -----BEGIN CERTIFICATE-----
    ...
    -----END CERTIFICATE-----

apiVersion: v1
kind: ConfigMap
metadata:
  name: my-search-external-ca
  namespace: mongodb
data:
  ca.crt: |
    -----BEGIN CERTIFICATE-----
    ...
    -----END CERTIFICATE-----

apiVersion: mongodb.com/v1
kind: MongoDBSearch
metadata:
  name: my-search
  namespace: mongodb
spec:
  source:
    external:
      hostAndPorts:
        - mongodb-0.example.com:27017
      tls:
        ca:
          name: my-search-external-ca

kubectl create configmap my-search-external-ca \
  --from-file=ca.crt=./ca.crt \
  --namespace mongodb

What Happens to Your Kubernetes Resources

This section covers what happens to the underlying Kubernetes objects that the Kubernetes Operator manages, and what to verify and clean up afterward. What changes depends on whether your MongoDBSearch resource syncs from a replica set or a sharded cluster.

kubectl delete statefulset my-search-search -n mongodb
kubectl delete service my-search-search-svc -n mongodb
kubectl delete configmap my-search-search-config -n mongodb

# Preview the PVCs to delete:
PVCS=$(kubectl get pvc -n mongodb -o name \
  | grep -E 'data-my-search-search-[0-9]+$')
echo "$PVCS"
# After confirming, delete exactly what you previewed:
echo "$PVCS" | xargs kubectl delete -n mongodb

Verification

Complete this checklist after you upgrade the Kubernetes Operator and apply the manifest changes, and before you delete any old resources:

• Confirm that kubectl get mongodbsearch <name> -n <namespace> -o yaml shows spec.clusters populated, and no leftover top-level replicas, persistence, resourceRequirements, statefulSet, loadBalancer, jvmFlags, or prometheus fields.

• Confirm that the new mongot StatefulSet, <name>-search-0 for a single-cluster deployment, has all pods Running and Ready.

• Confirm that mongot has finished reindexing on the new pods before you delete anything from the old StatefulSet. Check mongot's logs or index status for completion, not only pod readiness.

• If you rely on Prometheus scraping, confirm that your scrape configuration or dashboards reflect the new default-enabled behavior: either the metrics now appear, or your explicit mode: disabled setting keeps them disabled.

• If you use spec.source.external with TLS, confirm that the CA ConfigMap referenced by spec.source.external.tls.ca.name exists and contains a valid ca.crt key, and that mongot's connection to the external MongoDB deployment is healthy.

• For replica set sources, confirm that you've located and deleted the old <name>-search, <name>-search-svc, and <name>-search-config objects and their orphaned PVCs. For sharded sources, confirm that you've deleted the old per-shard TLS Secrets.

Get Support

If any part of your upgrade doesn't match what this guide describes, or you encounter an issue that this guide doesn't cover, contact MongoDB Support. Include your MongoDBSearch manifest with secrets redacted, the Kubernetes Operator version you're upgrading from and to, and the output of:

kubectl describe mongodbsearch <name> -n <namespace>

Learn More

• Search and Vector Search Compatibility and Requirements
• Upgrade or Downgrade
• MongoDBSearch Resource Specification
• mongot Deployment Frequently Asked Questions