Lvms:setup Prereq

---

name: lvms:setup-prereq argument-hint: "[connected|disconnected]" description: Set up prerequisites to test unreleased LVMS operator builds on OpenShift clusters user-invocable: true allowed-tools: Bash, Read, Write, Glob, Grep

lvms:setup-prereq

Synopsis

/lvms:setup-prereq [connected|disconnected]

Examples:

# Interactive (will prompt for deployment type)
/lvms:setup-prereq

# Connected cluster
/lvms:setup-prereq connected

# Disconnected cluster
/lvms:setup-prereq disconnected

Description

Sets up prerequisites to test unreleased LVMS operator builds on OpenShift clusters. Supports two deployment scenarios:

1. Connected cluster: Using CatalogSource and ImageDigestMirrorSet (IDMS)
2. Disconnected cluster: Using oc-mirror to mirror images to a disconnected registry

Required Information

Ask the user for:

1. Deployment type: Connected or Disconnected cluster
2. Kubeconfig path: Path to the kubeconfig file (default: ~/.kube/config)
3. If Connected: Catalog image (e.g., quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator-catalog@sha256:...)
4. If Disconnected:
   • OCP version (e.g., 4.17, 4.18, 4.19, 4.20)
   • Mirror registry URL (e.g., registry.example.com:5000)
   • Path to registry credentials file (default: ${XDG_RUNTIME_DIR}/containers/auth.json)
   • Catalog image to mirror

To get the catalog image, insert the snapshot name in: <konflux_prod_url>/ns/logical-volume-manag-tenant/applications/lvm-operator-catalog-<OCP_VERSION>/snapshots/<SNAPSHOT_NAME>

Implementation

Common Validation (Both Flows)

1. Check kubeconfig: ls -l <kubeconfig-path>
2. Verify oc available: which oc
3. Test connectivity: oc whoami --kubeconfig=<kubeconfig-path>
4. Validate catalog image format (must contain @sha256: or :tag)
5. Disconnected only: Verify oc-mirror: which oc-mirror
6. Disconnected only: Check credentials file exists
7. Disconnected only: Inform user to ensure mirror registry credentials are in their auth file. Credentials can be retrieved via:

   oc get secret pull-secret -n openshift-config -o jsonpath='{.data.\.dockerconfigjson}' --kubeconfig=<kubeconfig-path> | base64 -d
   

---

Flow A: Connected Cluster (IDMS + CatalogSource)

Step 1: Clean Up Existing Resources

oc delete catalogsource qe-app-registry -n openshift-marketplace --kubeconfig=<kubeconfig-path> --ignore-not-found=true
oc get imagecontentsourcepolicy --kubeconfig=<kubeconfig-path>
# If any ICSP exists, warn user and delete:
oc delete imagecontentsourcepolicy --all --kubeconfig=<kubeconfig-path>

Step 2: Create CatalogSource

Write /tmp/lvms-catalogsource.yaml:

apiVersion: operators.coreos.com/v1alpha1
kind: CatalogSource
metadata:
  name: lvms-custom-catalog
  namespace: openshift-marketplace
spec:
  displayName: konflux
  publisher: OpenShift QE
  sourceType: grpc
  updateStrategy:
    registryPoll:
      interval: 15m
  image: <CATALOG_IMAGE>

Step 3: Create ImageDigestMirrorSet

Write /tmp/lvms-idms.yaml:

apiVersion: config.openshift.io/v1
kind: ImageDigestMirrorSet
metadata:
  name: lvm-operator-imagedigestmirrors
spec:
  imageDigestMirrors:
    - mirrors:
      - registry.stage.redhat.io/lvms4/lvms-operator-bundle
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator-bundle
      source: registry.redhat.io/lvms4/lvms-operator-bundle
    - mirrors:
      - registry.stage.redhat.io/lvms4/lvms-rhel9-operator
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator
      source: registry.redhat.io/lvms4/lvms-rhel9-operator
    - mirrors:
      - registry.stage.redhat.io/lvms4/lvms-must-gather-rhel9
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvms-must-gather
      source: registry.redhat.io/lvms4/lvms-must-gather-rhel9
    - mirrors:
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator-bundle
      source: registry.stage.redhat.io/lvms4/lvms-operator-bundle
    - mirrors:
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator
      source: registry.stage.redhat.io/lvms4/lvms-rhel9-operator
    - mirrors:
      - quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvms-must-gather
      source: registry.stage.redhat.io/lvms4/lvms-must-gather-rhel9

Step 4: Apply IDMS and Monitor MCP

Warn: "Applying IDMS will trigger node reboots. This may take 20-30 minutes."

oc apply -f /tmp/lvms-idms.yaml --kubeconfig=<kubeconfig-path>
oc wait mcp/master mcp/worker --for=condition=Updating --timeout=5m --kubeconfig=<kubeconfig-path>
oc get mcp --kubeconfig=<kubeconfig-path>
oc wait mcp/master mcp/worker --for=condition=Updated --for=condition=Updating=False --timeout=30m --kubeconfig=<kubeconfig-path>
oc get mcp --kubeconfig=<kubeconfig-path>

Step 5: Apply CatalogSource

oc apply -f /tmp/lvms-catalogsource.yaml --kubeconfig=<kubeconfig-path>
oc wait catalogsource/lvms-custom-catalog -n openshift-marketplace --for=jsonpath='{.status.connectionState.lastObservedState}'=READY --timeout=5m --kubeconfig=<kubeconfig-path>

Step 6: Verify

oc get catalogsource lvms-custom-catalog -n openshift-marketplace --kubeconfig=<kubeconfig-path>
oc get packagemanifest lvms-operator --kubeconfig=<kubeconfig-path>

Inform: "Prerequisites are set up. You can now install the unreleased LVMS operator from OperatorHub or create a Subscription."

---

Flow B: Disconnected Cluster (oc-mirror)

Note: oc-mirror may not work reliably on macOS. Use Fedora or RHEL for best results.

Step 1: Create ImageSetConfiguration

Write /tmp/lvms-catalog-config.yaml:

kind: ImageSetConfiguration
apiVersion: mirror.openshift.io/v2alpha1
mirror:
  operators:
    - catalog: <CATALOG_IMAGE>
      packages:
        - name: lvms-operator
          channels:
            - name: 'stable-<OCP_VERSION>'

Step 2: Run oc-mirror for Catalog

mkdir -p /tmp/oc-mirror-workspace
oc-mirror -c /tmp/lvms-catalog-config.yaml --workspace file:///tmp/oc-mirror-workspace docker://<MIRROR_REGISTRY> --v2 --dest-tls-verify=false

Check for errors: ls -ltr /tmp/oc-mirror-workspace/working-dir/logs/mirroring_errors_*.txt

Step 3: Apply Generated CatalogSource

oc apply -f /tmp/oc-mirror-workspace/working-dir/cluster-resources/cs-*.yaml --kubeconfig=<kubeconfig-path>
oc wait catalogsource -n openshift-marketplace --all --for=jsonpath='{.status.connectionState.lastObservedState}'=READY --timeout=5m --kubeconfig=<kubeconfig-path>

Step 4: Handle Mirroring Errors (if any)

If mirroring errors exist, extract failed images and create corrected references:

• registry.redhat.io/lvms4/lvms-operator-bundle -> quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator-bundle
• registry.redhat.io/lvms4/lvms-rhel9-operator -> quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator
• registry.redhat.io/lvms4/lvms-must-gather-rhel9 -> quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvms-must-gather

Write /tmp/lvms-additional-images-config.yaml:

kind: ImageSetConfiguration
apiVersion: mirror.openshift.io/v2alpha1
mirror:
  additionalImages:
    - name: quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator-bundle@sha256:<SHA_FROM_LOG>
    - name: quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvm-operator@sha256:<SHA_FROM_LOG>
    - name: quay.io/redhat-user-workloads/logical-volume-manag-tenant/lvms-must-gather@sha256:<SHA_FROM_LOG>

Step 5: Mirror Additional Images (if Step 4 executed)

oc-mirror -c /tmp/lvms-additional-images-config.yaml --workspace file:///tmp/oc-mirror-workspace docker://<MIRROR_REGISTRY> --v2 --dest-tls-verify=false

Step 6: Create and Apply IDMS

Write /tmp/lvms-idms.yaml with mirrors pointing to <MIRROR_REGISTRY> (same structure as connected flow, with mirror registry as first mirror entry).

Step 7: Apply IDMS and Monitor MCP

Same as connected flow Step 4.

Step 8: Verify

oc get catalogsource -n openshift-marketplace --kubeconfig=<kubeconfig-path>
oc get packagemanifest lvms-operator --kubeconfig=<kubeconfig-path>

Inform: "Prerequisites are set up for disconnected cluster. You can now install the LVMS operator from OperatorHub or create a Subscription."

Important Notes

• ImageDigestMirrorSet triggers MachineConfigPool updates and node reboots
• Deleting ImageContentSourcePolicy also triggers MCP updates
• MCP updates can take 20-30 minutes depending on cluster size
• User must have cluster-admin permissions
• Always wait for MCP to start updating before waiting for completion (avoids false positives)

Error Handling

Error	Action
Kubeconfig not found	Ask for correct path
Invalid catalog image format	Ask for correction
oc not available	Install OpenShift CLI
oc whoami fails	Verify kubeconfig and cluster accessibility
MCP doesn't start updating (5m)	Check oc get mcp -o yaml
MCP update times out	Provide manual check: oc get mcp and oc get nodes
oc-mirror not available	Install from OpenShift mirror downloads
Registry credentials missing	Ask for path or create via podman login
oc-mirror auth errors	Verify credentials file and registry connectivity
oc-mirror mirroring fails	Check disk space and connectivity
CatalogSource fails to apply	Verify mirrored catalog image exists in mirror registry