1. Documentation
×

Updating managed clusters with the Topology Aware Lifecycle Manager

About the Topology Aware Lifecycle Manager configuration

The Topology Aware Lifecycle Manager (TALM) manages the deployment of Red Hat Advanced Cluster Management (RHACM) policies for one or more OKD clusters. Using TALM in a large network of clusters allows the phased rollout of policies to the clusters in limited batches. This helps to minimize possible service disruptions when updating. With TALM, you can control the following actions:

  • The timing of the update

  • The number of RHACM-managed clusters

  • The subset of managed clusters to apply the policies to

  • The update order of the clusters

  • The set of policies remediated to the cluster

  • The order of policies remediated to the cluster

  • The assignment of a canary cluster

For single-node OpenShift, the Topology Aware Lifecycle Manager (TALM) offers the following features:

  • Create a backup of a deployment before an upgrade

  • Pre-caching images for clusters with limited bandwidth

TALM supports the orchestration of the OKD y-stream and z-stream updates, and day-two operations on y-streams and z-streams.

About managed policies used with Topology Aware Lifecycle Manager

The Topology Aware Lifecycle Manager (TALM) uses RHACM policies for cluster updates.

TALM can be used to manage the rollout of any policy CR where the remediationAction field is set to inform. Supported use cases include the following:

  • Manual user creation of policy CRs

  • Automatically generated policies from the PolicyGenTemplate custom resource definition (CRD)

For policies that update an Operator subscription with manual approval, TALM provides additional functionality that approves the installation of the updated Operator.

For more information about managed policies, see Policy Overview in the RHACM documentation.

For more information about the PolicyGenTemplate CRD, see the "About the PolicyGenTemplate CRD" section in "Configuring managed clusters with policies and PolicyGenTemplate resources".

Installing the Topology Aware Lifecycle Manager by using the web console

You can use the OKD web console to install the Topology Aware Lifecycle Manager.

Prerequisites

  • Install the latest version of the RHACM Operator.

  • Set up a hub cluster with disconnected regitry.

  • Log in as a user with cluster-admin privileges.

Procedure

  1. In the OKD web console, navigate to Operators → OperatorHub.

  2. Search for the Topology Aware Lifecycle Manager from the list of available Operators, and then click Install.

  3. Keep the default selection of Installation mode ["All namespaces on the cluster (default)"] and Installed Namespace ("openshift-operators") to ensure that the Operator is installed properly.

  4. Click Install.

Verification

To confirm that the installation is successful:

  1. Navigate to the Operators → Installed Operators page.

  2. Check that the Operator is installed in the All Namespaces namespace and its status is Succeeded.

If the Operator is not installed successfully:

  1. Navigate to the Operators → Installed Operators page and inspect the Status column for any errors or failures.

  2. Navigate to the Workloads → Pods page and check the logs in any containers in the cluster-group-upgrades-controller-manager pod that are reporting issues.

Installing the Topology Aware Lifecycle Manager by using the CLI

You can use the OpenShift CLI (oc) to install the Topology Aware Lifecycle Manager (TALM).

Prerequisites

  • Install the OpenShift CLI (oc).

  • Install the latest version of the RHACM Operator.

  • Set up a hub cluster with disconnected registry.

  • Log in as a user with cluster-admin privileges.

Procedure

  1. Create a Subscription CR:

    1. Define the Subscription CR and save the YAML file, for example, talm-subscription.yaml:

      apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
  name: openshift-topology-aware-lifecycle-manager-subscription
  namespace: openshift-operators
spec:
  channel: "stable"
  name: topology-aware-lifecycle-manager
  source: redhat-operators
  sourceNamespace: openshift-marketplace

    2. Create the Subscription CR by running the following command:

      $ oc create -f talm-subscription.yaml
Verification

  1. Verify that the installation succeeded by inspecting the CSV resource:

    $ oc get csv -n openshift-operators
    Example output
    NAME                                                   DISPLAY                            VERSION               REPLACES                           PHASE
topology-aware-lifecycle-manager.4.14.x   Topology Aware Lifecycle Manager   4.14.x                                      Succeeded

  2. Verify that the TALM is up and running:

    $ oc get deploy -n openshift-operators
    Example output
    NAMESPACE                                          NAME                                             READY   UP-TO-DATE   AVAILABLE   AGE
openshift-operators                                cluster-group-upgrades-controller-manager        1/1     1            1           14s

About the ClusterGroupUpgrade CR

The Topology Aware Lifecycle Manager (TALM) builds the remediation plan from the ClusterGroupUpgrade CR for a group of clusters. You can define the following specifications in a ClusterGroupUpgrade CR:

  • Clusters in the group

  • Blocking ClusterGroupUpgrade CRs

  • Applicable list of managed policies

  • Number of concurrent updates

  • Applicable canary updates

  • Actions to perform before and after the update

  • Update timing

You can control the start time of an update using the enable field in the ClusterGroupUpgrade CR. For example, if you have a scheduled maintenance window of four hours, you can prepare a ClusterGroupUpgrade CR with the enable field set to false.

You can set the timeout by configuring the spec.remediationStrategy.timeout setting as follows:

spec
  remediationStrategy:
          maxConcurrency: 1
          timeout: 240

You can use the batchTimeoutAction to determine what happens if an update fails for a cluster. You can specify continue to skip the failing cluster and continue to upgrade other clusters, or abort to stop policy remediation for all clusters. Once the timeout elapses, TALM removes all enforce policies to ensure that no further updates are made to clusters.

To apply the changes, you set the enabled field to true.

For more information see the "Applying update policies to managed clusters" section.

As TALM works through remediation of the policies to the specified clusters, the ClusterGroupUpgrade CR can report true or false statuses for a number of conditions.

After TALM completes a cluster update, the cluster does not update again under the control of the same ClusterGroupUpgrade CR. You must create a new ClusterGroupUpgrade CR in the following cases:

  • When you need to update the cluster again

  • When the cluster changes to non-compliant with the inform policy after being updated

Selecting clusters

TALM builds a remediation plan and selects clusters based on the following fields:

  • The clusterLabelSelector field specifies the labels of the clusters that you want to update. This consists of a list of the standard label selectors from k8s.io/apimachinery/pkg/apis/meta/v1. Each selector in the list uses either label value pairs or label expressions. Matches from each selector are added to the final list of clusters along with the matches from the clusterSelector field and the cluster field.

  • The clusters field specifies a list of clusters to update.

  • The canaries field specifies the clusters for canary updates.

  • The maxConcurrency field specifies the number of clusters to update in a batch.

  • The actions field specifies beforeEnable actions that TALM takes as it begins the update process, and afterCompletion actions that TALM takes as it completes policy remediation for each cluster.

You can use the clusters, clusterLabelSelector, and clusterSelector fields together to create a combined list of clusters.

The remediation plan starts with the clusters listed in the canaries field. Each canary cluster forms a single-cluster batch.

Sample ClusterGroupUpgrade CR with the enabled field set to false
apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  creationTimestamp: '2022-11-18T16:27:15Z'
  finalizers:
    - ran.openshift.io/cleanup-finalizer
  generation: 1
  name: talm-cgu
  namespace: talm-namespace
  resourceVersion: '40451823'
  uid: cca245a5-4bca-45fa-89c0-aa6af81a596c
Spec:
  actions:
    afterCompletion: (1)
      addClusterLabels:
        upgrade-done: ""
      deleteClusterLabels:
        upgrade-running: ""
      deleteObjects: true
    beforeEnable: (2)
      addClusterLabels:
        upgrade-running: ""
  backup: false
  clusters: (3)
    - spoke1
  enable: false (4)
  managedPolicies: (5)
    - talm-policy
  preCaching: false
  remediationStrategy: (6)
    canaries: (7)
        - spoke1
    maxConcurrency: 2 (8)
    timeout: 240
  clusterLabelSelectors: (9)
    - matchExpressions:
      - key: label1
      operator: In
      values:
        - value1a
        - value1b
  batchTimeoutAction: (10)
status: (11)
    computedMaxConcurrency: 2
    conditions:
      - lastTransitionTime: '2022-11-18T16:27:15Z'
        message: All selected clusters are valid
        reason: ClusterSelectionCompleted
        status: 'True'
        type: ClustersSelected (12)
      - lastTransitionTime: '2022-11-18T16:27:15Z'
        message: Completed validation
        reason: ValidationCompleted
        status: 'True'
        type: Validated (13)
      - lastTransitionTime: '2022-11-18T16:37:16Z'
        message: Not enabled
        reason: NotEnabled
        status: 'False'
        type: Progressing
    managedPoliciesForUpgrade:
      - name: talm-policy
        namespace: talm-namespace
    managedPoliciesNs:
      talm-policy: talm-namespace
    remediationPlan:
      - - spoke1
      - - spoke2
        - spoke3
    status:
1 Specifies the action that TALM takes when it completes policy remediation for each cluster.
2 Specifies the action that TALM takes as it begins the update process.
3 Defines the list of clusters to update.
4 The enable field is set to false.
5 Lists the user-defined set of policies to remediate.
6 Defines the specifics of the cluster updates.
7 Defines the clusters for canary updates.
8 Defines the maximum number of concurrent updates in a batch. The number of remediation batches is the number of canary clusters, plus the number of clusters, except the canary clusters, divided by the maxConcurrency value. The clusters that are already compliant with all the managed policies are excluded from the remediation plan.
9 Displays the parameters for selecting clusters.
10 Controls what happens if a batch times out. Possible values are abort or continue. If unspecified, the default is continue.
11 Displays information about the status of the updates.
12 The ClustersSelected condition shows that all selected clusters are valid.
13 The Validated condition shows that all selected clusters have been validated.

Any failures during the update of a canary cluster stops the update process.

When the remediation plan is successfully created, you can you set the enable field to true and TALM starts to update the non-compliant clusters with the specified managed policies.

You can only make changes to the spec fields if the enable field of the ClusterGroupUpgrade CR is set to false.

Validating

TALM checks that all specified managed policies are available and correct, and uses the Validated condition to report the status and reasons as follows:

  • true

    Validation is completed.

  • false

    Policies are missing or invalid, or an invalid platform image has been specified.

Pre-caching

Clusters might have limited bandwidth to access the container image registry, which can cause a timeout before the updates are completed. On single-node OpenShift clusters, you can use pre-caching to avoid this. The container image pre-caching starts when you create a ClusterGroupUpgrade CR with the preCaching field set to true. TALM compares the available disk space with the estimated OKD image size to ensure that there is enough space. If a cluster has insufficient space, TALM cancels pre-caching for that cluster and does not remediate policies on it.

TALM uses the PrecacheSpecValid condition to report status information as follows:

  • true

    The pre-caching spec is valid and consistent.

  • false

    The pre-caching spec is incomplete.

TALM uses the PrecachingSucceeded condition to report status information as follows:

  • true

    TALM has concluded the pre-caching process. If pre-caching fails for any cluster, the update fails for that cluster but proceeds for all other clusters. A message informs you if pre-caching has failed for any clusters.

  • false

    Pre-caching is still in progress for one or more clusters or has failed for all clusters.

For more information see the "Using the container image pre-cache feature" section.

Creating a backup

For single-node OpenShift, TALM can create a backup of a deployment before an update. If the update fails, you can recover the previous version and restore a cluster to a working state without requiring a reprovision of applications. To use the backup feature you first create a ClusterGroupUpgrade CR with the backup field set to true. To ensure that the contents of the backup are up to date, the backup is not taken until you set the enable field in the ClusterGroupUpgrade CR to true.

TALM uses the BackupSucceeded condition to report the status and reasons as follows:

  • true

    Backup is completed for all clusters or the backup run has completed but failed for one or more clusters. If backup fails for any cluster, the update fails for that cluster but proceeds for all other clusters.

  • false

    Backup is still in progress for one or more clusters or has failed for all clusters.

For more information, see the "Creating a backup of cluster resources before upgrade" section.

Updating clusters

TALM enforces the policies following the remediation plan. Enforcing the policies for subsequent batches starts immediately after all the clusters of the current batch are compliant with all the managed policies. If the batch times out, TALM moves on to the next batch. The timeout value of a batch is the spec.timeout field divided by the number of batches in the remediation plan.

TALM uses the Progressing condition to report the status and reasons as follows:

  • true

    TALM is remediating non-compliant policies.

  • false

    The update is not in progress. Possible reasons for this are:

    • All clusters are compliant with all the managed policies.

    • The update has timed out as policy remediation took too long.

    • Blocking CRs are missing from the system or have not yet completed.

    • The ClusterGroupUpgrade CR is not enabled.

    • Backup is still in progress.

The managed policies apply in the order that they are listed in the managedPolicies field in the ClusterGroupUpgrade CR. One managed policy is applied to the specified clusters at a time. When a cluster complies with the current policy, the next managed policy is applied to it.
Sample ClusterGroupUpgrade CR in the Progressing state
apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  creationTimestamp: '2022-11-18T16:27:15Z'
  finalizers:
    - ran.openshift.io/cleanup-finalizer
  generation: 1
  name: talm-cgu
  namespace: talm-namespace
  resourceVersion: '40451823'
  uid: cca245a5-4bca-45fa-89c0-aa6af81a596c
Spec:
  actions:
    afterCompletion:
      deleteObjects: true
    beforeEnable: {}
  backup: false
  clusters:
    - spoke1
  enable: true
  managedPolicies:
    - talm-policy
  preCaching: true
  remediationStrategy:
    canaries:
        - spoke1
    maxConcurrency: 2
    timeout: 240
  clusterLabelSelectors:
    - matchExpressions:
      - key: label1
      operator: In
      values:
        - value1a
        - value1b
  batchTimeoutAction:
status:
    clusters:
      - name: spoke1
        state: complete
    computedMaxConcurrency: 2
    conditions:
      - lastTransitionTime: '2022-11-18T16:27:15Z'
        message: All selected clusters are valid
        reason: ClusterSelectionCompleted
        status: 'True'
        type: ClustersSelected
      - lastTransitionTime: '2022-11-18T16:27:15Z'
        message: Completed validation
        reason: ValidationCompleted
        status: 'True'
        type: Validated
      - lastTransitionTime: '2022-11-18T16:37:16Z'
        message: Remediating non-compliant policies
        reason: InProgress
        status: 'True'
        type: Progressing (1)
    managedPoliciesForUpgrade:
      - name: talm-policy
        namespace: talm-namespace
    managedPoliciesNs:
      talm-policy: talm-namespace
    remediationPlan:
      - - spoke1
      - - spoke2
        - spoke3
    status:
      currentBatch: 2
      currentBatchRemediationProgress:
        spoke2:
          state: Completed
        spoke3:
          policyIndex: 0
          state: InProgress
      currentBatchStartedAt: '2022-11-18T16:27:16Z'
      startedAt: '2022-11-18T16:27:15Z'
1 The Progressing fields show that TALM is in the process of remediating policies.

Update status

TALM uses the Succeeded condition to report the status and reasons as follows:

  • true

    All clusters are compliant with the specified managed policies.

  • false

    Policy remediation failed as there were no clusters available for remediation, or because policy remediation took too long for one of the following reasons:

    • The current batch contains canary updates and the cluster in the batch does not comply with all the managed policies within the batch timeout.

    • Clusters did not comply with the managed policies within the timeout value specified in the remediationStrategy field.

Sample ClusterGroupUpgrade CR in the Succeeded state
    apiVersion: ran.openshift.io/v1alpha1
    kind: ClusterGroupUpgrade
    metadata:
      name: cgu-upgrade-complete
      namespace: default
    spec:
      clusters:
      - spoke1
      - spoke4
      enable: true
      managedPolicies:
      - policy1-common-cluster-version-policy
      - policy2-common-pao-sub-policy
      remediationStrategy:
        maxConcurrency: 1
        timeout: 240
    status: (3)
      clusters:
        - name: spoke1
          state: complete
        - name: spoke4
          state: complete
      conditions:
      - message: All selected clusters are valid
        reason: ClusterSelectionCompleted
        status: "True"
        type: ClustersSelected
      - message: Completed validation
        reason: ValidationCompleted
        status: "True"
        type: Validated
      - message: All clusters are compliant with all the managed policies
        reason: Completed
        status: "False"
        type: Progressing (1)
      - message: All clusters are compliant with all the managed policies
        reason: Completed
        status: "True"
        type: Succeeded (2)
      managedPoliciesForUpgrade:
      - name: policy1-common-cluster-version-policy
        namespace: default
      - name: policy2-common-pao-sub-policy
        namespace: default
      remediationPlan:
      - - spoke1
      - - spoke4
      status:
        completedAt: '2022-11-18T16:27:16Z'
        startedAt: '2022-11-18T16:27:15Z'
1 In the Progressing fields, the status is false as the update has completed; clusters are compliant with all the managed policies.
2 The Succeeded fields show that the validations completed successfully.
3 The status field includes a list of clusters and their respective statuses. The status of a cluster can be complete or timedout.
Sample ClusterGroupUpgrade CR in the timedout state
apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  creationTimestamp: '2022-11-18T16:27:15Z'
  finalizers:
    - ran.openshift.io/cleanup-finalizer
  generation: 1
  name: talm-cgu
  namespace: talm-namespace
  resourceVersion: '40451823'
  uid: cca245a5-4bca-45fa-89c0-aa6af81a596c
spec:
  actions:
    afterCompletion:
      deleteObjects: true
    beforeEnable: {}
  backup: false
  clusters:
    - spoke1
    - spoke2
  enable: true
  managedPolicies:
    - talm-policy
  preCaching: false
  remediationStrategy:
    maxConcurrency: 2
    timeout: 240
status:
  clusters:
    - name: spoke1
      state: complete
    - currentPolicy: (1)
        name: talm-policy
        status: NonCompliant
      name: spoke2
      state: timedout
  computedMaxConcurrency: 2
  conditions:
    - lastTransitionTime: '2022-11-18T16:27:15Z'
      message: All selected clusters are valid
      reason: ClusterSelectionCompleted
      status: 'True'
      type: ClustersSelected
    - lastTransitionTime: '2022-11-18T16:27:15Z'
      message: Completed validation
      reason: ValidationCompleted
      status: 'True'
      type: Validated
    - lastTransitionTime: '2022-11-18T16:37:16Z'
      message: Policy remediation took too long
      reason: TimedOut
      status: 'False'
      type: Progressing
    - lastTransitionTime: '2022-11-18T16:37:16Z'
      message: Policy remediation took too long
      reason: TimedOut
      status: 'False'
      type: Succeeded (2)
  managedPoliciesForUpgrade:
    - name: talm-policy
      namespace: talm-namespace
  managedPoliciesNs:
    talm-policy: talm-namespace
  remediationPlan:
    - - spoke1
      - spoke2
  status:
        startedAt: '2022-11-18T16:27:15Z'
        completedAt: '2022-11-18T20:27:15Z'
1 If a cluster’s state is timedout, the currentPolicy field shows the name of the policy and the policy status.
2 The status for succeeded is false and the message indicates that policy remediation took too long.

Blocking ClusterGroupUpgrade CRs

You can create multiple ClusterGroupUpgrade CRs and control their order of application.

For example, if you create ClusterGroupUpgrade CR C that blocks the start of ClusterGroupUpgrade CR A, then ClusterGroupUpgrade CR A cannot start until the status of ClusterGroupUpgrade CR C becomes UpgradeComplete.

One ClusterGroupUpgrade CR can have multiple blocking CRs. In this case, all the blocking CRs must complete before the upgrade for the current CR can start.

Prerequisites

  • Install the Topology Aware Lifecycle Manager (TALM).

  • Provision one or more managed clusters.

  • Log in as a user with cluster-admin privileges.

  • Create RHACM policies in the hub cluster.

Procedure

  1. Save the content of the ClusterGroupUpgrade CRs in the cgu-a.yaml, cgu-b.yaml, and cgu-c.yaml files.

    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-a
  namespace: default
spec:
  blockingCRs: (1)
  - name: cgu-c
    namespace: default
  clusters:
  - spoke1
  - spoke2
  - spoke3
  enable: false
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  remediationStrategy:
    canaries:
    - spoke1
    maxConcurrency: 2
    timeout: 240
status:
  conditions:
  - message: The ClusterGroupUpgrade CR is not enabled
    reason: UpgradeNotStarted
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy2-common-pao-sub-policy
    namespace: default
  - name: policy3-common-ptp-sub-policy
    namespace: default
  placementBindings:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  placementRules:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  remediationPlan:
  - - spoke1
  - - spoke2
    1 Defines the blocking CRs. The cgu-a update cannot start until cgu-c is complete. 
    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-b
  namespace: default
spec:
  blockingCRs: (1)
  - name: cgu-a
    namespace: default
  clusters:
  - spoke4
  - spoke5
  enable: false
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  - policy4-common-sriov-sub-policy
  remediationStrategy:
    maxConcurrency: 1
    timeout: 240
status:
  conditions:
  - message: The ClusterGroupUpgrade CR is not enabled
    reason: UpgradeNotStarted
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy2-common-pao-sub-policy
    namespace: default
  - name: policy3-common-ptp-sub-policy
    namespace: default
  - name: policy4-common-sriov-sub-policy
    namespace: default
  placementBindings:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  placementRules:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  remediationPlan:
  - - spoke4
  - - spoke5
  status: {}
    1 The cgu-b update cannot start until cgu-a is complete. 
    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-c
  namespace: default
spec: (1)
  clusters:
  - spoke6
  enable: false
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  - policy4-common-sriov-sub-policy
  remediationStrategy:
    maxConcurrency: 1
    timeout: 240
status:
  conditions:
  - message: The ClusterGroupUpgrade CR is not enabled
    reason: UpgradeNotStarted
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  managedPoliciesCompliantBeforeUpgrade:
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy4-common-sriov-sub-policy
    namespace: default
  placementBindings:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  placementRules:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  remediationPlan:
  - - spoke6
  status: {}
    1 The cgu-c update does not have any blocking CRs. TALM starts the cgu-c update when the enable field is set to true.

  2. Create the ClusterGroupUpgrade CRs by running the following command for each relevant CR:

    $ oc apply -f <name>.yaml

  3. Start the update process by running the following command for each relevant CR:

    $ oc --namespace=default patch clustergroupupgrade.ran.openshift.io/<name> \
--type merge -p '{"spec":{"enable":true}}'

    The following examples show ClusterGroupUpgrade CRs where the enable field is set to true:

    Example for cgu-a with blocking CRs
    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-a
  namespace: default
spec:
  blockingCRs:
  - name: cgu-c
    namespace: default
  clusters:
  - spoke1
  - spoke2
  - spoke3
  enable: true
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  remediationStrategy:
    canaries:
    - spoke1
    maxConcurrency: 2
    timeout: 240
status:
  conditions:
  - message: 'The ClusterGroupUpgrade CR is blocked by other CRs that have not yet
      completed: [cgu-c]' (1)
    reason: UpgradeCannotStart
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy2-common-pao-sub-policy
    namespace: default
  - name: policy3-common-ptp-sub-policy
    namespace: default
  placementBindings:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  placementRules:
  - cgu-a-policy1-common-cluster-version-policy
  - cgu-a-policy2-common-pao-sub-policy
  - cgu-a-policy3-common-ptp-sub-policy
  remediationPlan:
  - - spoke1
  - - spoke2
  status: {}
    1 Shows the list of blocking CRs.
    Example for cgu-b with blocking CRs
    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-b
  namespace: default
spec:
  blockingCRs:
  - name: cgu-a
    namespace: default
  clusters:
  - spoke4
  - spoke5
  enable: true
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  - policy4-common-sriov-sub-policy
  remediationStrategy:
    maxConcurrency: 1
    timeout: 240
status:
  conditions:
  - message: 'The ClusterGroupUpgrade CR is blocked by other CRs that have not yet
      completed: [cgu-a]' (1)
    reason: UpgradeCannotStart
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy2-common-pao-sub-policy
    namespace: default
  - name: policy3-common-ptp-sub-policy
    namespace: default
  - name: policy4-common-sriov-sub-policy
    namespace: default
  placementBindings:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  placementRules:
  - cgu-b-policy1-common-cluster-version-policy
  - cgu-b-policy2-common-pao-sub-policy
  - cgu-b-policy3-common-ptp-sub-policy
  - cgu-b-policy4-common-sriov-sub-policy
  remediationPlan:
  - - spoke4
  - - spoke5
  status: {}
    1 Shows the list of blocking CRs.
    Example for cgu-c with blocking CRs
    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-c
  namespace: default
spec:
  clusters:
  - spoke6
  enable: true
  managedPolicies:
  - policy1-common-cluster-version-policy
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  - policy4-common-sriov-sub-policy
  remediationStrategy:
    maxConcurrency: 1
    timeout: 240
status:
  conditions:
  - message: The ClusterGroupUpgrade CR has upgrade policies that are still non compliant (1)
    reason: UpgradeNotCompleted
    status: "False"
    type: Ready
  copiedPolicies:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  managedPoliciesCompliantBeforeUpgrade:
  - policy2-common-pao-sub-policy
  - policy3-common-ptp-sub-policy
  managedPoliciesForUpgrade:
  - name: policy1-common-cluster-version-policy
    namespace: default
  - name: policy4-common-sriov-sub-policy
    namespace: default
  placementBindings:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  placementRules:
  - cgu-c-policy1-common-cluster-version-policy
  - cgu-c-policy4-common-sriov-sub-policy
  remediationPlan:
  - - spoke6
  status:
    currentBatch: 1
    remediationPlanForBatch:
      spoke6: 0
    1 The cgu-c update does not have any blocking CRs.

Update policies on managed clusters

The Topology Aware Lifecycle Manager (TALM) remediates a set of inform policies for the clusters specified in the ClusterGroupUpgrade CR. TALM remediates inform policies by making enforce copies of the managed RHACM policies. Each copied policy has its own corresponding RHACM placement rule and RHACM placement binding.

One by one, TALM adds each cluster from the current batch to the placement rule that corresponds with the applicable managed policy. If a cluster is already compliant with a policy, TALM skips applying that policy on the compliant cluster. TALM then moves on to applying the next policy to the non-compliant cluster. After TALM completes the updates in a batch, all clusters are removed from the placement rules associated with the copied policies. Then, the update of the next batch starts.

If a spoke cluster does not report any compliant state to RHACM, the managed policies on the hub cluster can be missing status information that TALM needs. TALM handles these cases in the following ways:

  • If a policy’s status.compliant field is missing, TALM ignores the policy and adds a log entry. Then, TALM continues looking at the policy’s status.status field.

  • If a policy’s status.status is missing, TALM produces an error.

  • If a cluster’s compliance status is missing in the policy’s status.status field, TALM considers that cluster to be non-compliant with that policy.

The ClusterGroupUpgrade CR’s batchTimeoutAction determines what happens if an upgrade fails for a cluster. You can specify continue to skip the failing cluster and continue to upgrade other clusters, or specify abort to stop the policy remediation for all clusters. Once the timeout elapses, TALM removes all enforce policies to ensure that no further updates are made to clusters.

Example upgrade policy
apiVersion: policy.open-cluster-management.io/v1
kind: Policy
metadata:
  name: ocp-4.4.14.4
  namespace: platform-upgrade
spec:
  disabled: false
  policy-templates:
  - objectDefinition:
      apiVersion: policy.open-cluster-management.io/v1
      kind: ConfigurationPolicy
      metadata:
        name: upgrade
      spec:
        namespaceselector:
          exclude:
          - kube-*
          include:
          - '*'
        object-templates:
        - complianceType: musthave
          objectDefinition:
            apiVersion: config.openshift.io/v1
            kind: ClusterVersion
            metadata:
              name: version
            spec:
              channel: stable-4.14
              desiredUpdate:
                version: 4.4.14.4
              upstream: https://api.openshift.com/api/upgrades_info/v1/graph
            status:
              history:
                - state: Completed
                  version: 4.4.14.4
        remediationAction: inform
        severity: low
  remediationAction: inform

For more information about RHACM policies, see Policy overview.

Additional resources

For more information about the PolicyGenTemplate CRD, see About the PolicyGenTemplate CRD.

Configuring Operator subscriptions for managed clusters that you install with TALM

Topology Aware Lifecycle Manager (TALM) can only approve the install plan for an Operator if the Subscription custom resource (CR) of the Operator contains the status.state.AtLatestKnown field.

Procedure

  1. Add the status.state.AtLatestKnown field to the Subscription CR of the Operator:

    Example Subscription CR
    apiVersion: operators.coreos.com/v1alpha1
kind: Subscription
metadata:
  name: cluster-logging
  namespace: openshift-logging
  annotations:
    ran.openshift.io/ztp-deploy-wave: "2"
spec:
  channel: "stable"
  name: cluster-logging
  source: redhat-operators
  sourceNamespace: openshift-marketplace
  installPlanApproval: Manual
status:
  state: AtLatestKnown (1)
    1 The status.state: AtLatestKnown field is used for the latest Operator version available from the Operator catalog.

    When a new version of the Operator is available in the registry, the associated policy becomes non-compliant.

  2. Apply the changed Subscription policy to your managed clusters with a ClusterGroupUpgrade CR.

Applying update policies to managed clusters

You can update your managed clusters by applying your policies.

Prerequisites

  • Install the Topology Aware Lifecycle Manager (TALM).

  • Provision one or more managed clusters.

  • Log in as a user with cluster-admin privileges.

  • Create RHACM policies in the hub cluster.

Procedure

  1. Save the contents of the ClusterGroupUpgrade CR in the cgu-1.yaml file.

    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: cgu-1
  namespace: default
spec:
  managedPolicies: (1)
    - policy1-common-cluster-version-policy
    - policy2-common-nto-sub-policy
    - policy3-common-ptp-sub-policy
    - policy4-common-sriov-sub-policy
  enable: false
  clusters: (2)
  - spoke1
  - spoke2
  - spoke5
  - spoke6
  remediationStrategy:
    maxConcurrency: 2 (3)
    timeout: 240 (4)
  batchTimeoutAction: (5)
    1 The name of the policies to apply.
    2 The list of clusters to update.
    3 The maxConcurrency field signifies the number of clusters updated at the same time.
    4 The update timeout in minutes.
    5 Controls what happens if a batch times out. Possible values are abort or continue. If unspecified, the default is continue.

  2. Create the ClusterGroupUpgrade CR by running the following command:

    $ oc create -f cgu-1.yaml

    1. Check if the ClusterGroupUpgrade CR was created in the hub cluster by running the following command:

      $ oc get cgu --all-namespaces
      Example output
      NAMESPACE   NAME  AGE  STATE      DETAILS
default     cgu-1 8m55 NotEnabled Not Enabled

    2. Check the status of the update by running the following command:

      $ oc get cgu -n default cgu-1 -ojsonpath='{.status}' | jq
      Example output
      {
  "computedMaxConcurrency": 2,
  "conditions": [
    {
      "lastTransitionTime": "2022-02-25T15:34:07Z",
      "message": "Not enabled", (1)
      "reason": "NotEnabled",
      "status": "False",
      "type": "Progressing"
    }
  ],
  "copiedPolicies": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "managedPoliciesContent": {
    "policy1-common-cluster-version-policy": "null",
    "policy2-common-nto-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"node-tuning-operator\",\"namespace\":\"openshift-cluster-node-tuning-operator\"}]",
    "policy3-common-ptp-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"ptp-operator-subscription\",\"namespace\":\"openshift-ptp\"}]",
    "policy4-common-sriov-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"sriov-network-operator-subscription\",\"namespace\":\"openshift-sriov-network-operator\"}]"
  },
  "managedPoliciesForUpgrade": [
    {
      "name": "policy1-common-cluster-version-policy",
      "namespace": "default"
    },
    {
      "name": "policy2-common-nto-sub-policy",
      "namespace": "default"
    },
    {
      "name": "policy3-common-ptp-sub-policy",
      "namespace": "default"
    },
    {
      "name": "policy4-common-sriov-sub-policy",
      "namespace": "default"
    }
  ],
  "managedPoliciesNs": {
    "policy1-common-cluster-version-policy": "default",
    "policy2-common-nto-sub-policy": "default",
    "policy3-common-ptp-sub-policy": "default",
    "policy4-common-sriov-sub-policy": "default"
  },
  "placementBindings": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "placementRules": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "precaching": {
    "spec": {}
  },
  "remediationPlan": [
    [
      "spoke1",
      "spoke2"
    ],
    [
      "spoke5",
      "spoke6"
    ]
  ],
  "status": {}
}
      1 The spec.enable field in the ClusterGroupUpgrade CR is set to false.

    3. Check the status of the policies by running the following command:

      $ oc get policies -A
      Example output
      NAMESPACE   NAME                                                 REMEDIATION ACTION   COMPLIANCE STATE   AGE
default     cgu-policy1-common-cluster-version-policy            enforce                                 17m (1)
default     cgu-policy2-common-nto-sub-policy                    enforce                                 17m
default     cgu-policy3-common-ptp-sub-policy                    enforce                                 17m
default     cgu-policy4-common-sriov-sub-policy                  enforce                                 17m
default     policy1-common-cluster-version-policy                inform               NonCompliant       15h
default     policy2-common-nto-sub-policy                        inform               NonCompliant       15h
default     policy3-common-ptp-sub-policy                        inform               NonCompliant       18m
default     policy4-common-sriov-sub-policy                      inform               NonCompliant       18m
      1 The spec.remediationAction field of policies currently applied on the clusters is set to enforce. The managed policies in inform mode from the ClusterGroupUpgrade CR remain in inform mode during the update.

  3. Change the value of the spec.enable field to true by running the following command:

    $ oc --namespace=default patch clustergroupupgrade.ran.openshift.io/cgu-1 \
--patch '{"spec":{"enable":true}}' --type=merge
Verification

  1. Check the status of the update again by running the following command:

    $ oc get cgu -n default cgu-1 -ojsonpath='{.status}' | jq
    Example output
    {
  "computedMaxConcurrency": 2,
  "conditions": [ (1)
    {
      "lastTransitionTime": "2022-02-25T15:33:07Z",
      "message": "All selected clusters are valid",
      "reason": "ClusterSelectionCompleted",
      "status": "True",
      "type": "ClustersSelected",
      "lastTransitionTime": "2022-02-25T15:33:07Z",
      "message": "Completed validation",
      "reason": "ValidationCompleted",
      "status": "True",
      "type": "Validated",
      "lastTransitionTime": "2022-02-25T15:34:07Z",
      "message": "Remediating non-compliant policies",
      "reason": "InProgress",
      "status": "True",
      "type": "Progressing"
    }
  ],
  "copiedPolicies": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "managedPoliciesContent": {
    "policy1-common-cluster-version-policy": "null",
    "policy2-common-nto-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"node-tuning-operator\",\"namespace\":\"openshift-cluster-node-tuning-operator\"}]",
    "policy3-common-ptp-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"ptp-operator-subscription\",\"namespace\":\"openshift-ptp\"}]",
    "policy4-common-sriov-sub-policy": "[{\"kind\":\"Subscription\",\"name\":\"sriov-network-operator-subscription\",\"namespace\":\"openshift-sriov-network-operator\"}]"
  },
  "managedPoliciesForUpgrade": [
    {
      "name": "policy1-common-cluster-version-policy",
      "namespace": "default"
    },
    {
      "name": "policy2-common-nto-sub-policy",
      "namespace": "default"
    },
    {
      "name": "policy3-common-ptp-sub-policy",
      "namespace": "default"
    },
    {
      "name": "policy4-common-sriov-sub-policy",
      "namespace": "default"
    }
  ],
  "managedPoliciesNs": {
    "policy1-common-cluster-version-policy": "default",
    "policy2-common-nto-sub-policy": "default",
    "policy3-common-ptp-sub-policy": "default",
    "policy4-common-sriov-sub-policy": "default"
  },
  "placementBindings": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "placementRules": [
    "cgu-policy1-common-cluster-version-policy",
    "cgu-policy2-common-nto-sub-policy",
    "cgu-policy3-common-ptp-sub-policy",
    "cgu-policy4-common-sriov-sub-policy"
  ],
  "precaching": {
    "spec": {}
  },
  "remediationPlan": [
    [
      "spoke1",
      "spoke2"
    ],
    [
      "spoke5",
      "spoke6"
    ]
  ],
  "status": {
    "currentBatch": 1,
    "currentBatchStartedAt": "2022-02-25T15:54:16Z",
    "remediationPlanForBatch": {
      "spoke1": 0,
      "spoke2": 1
    },
    "startedAt": "2022-02-25T15:54:16Z"
  }
}
    1 Reflects the update progress of the current batch. Run this command again to receive updated information about the progress.

  2. If the policies include Operator subscriptions, you can check the installation progress directly on the single-node cluster.

    1. Export the KUBECONFIG file of the single-node cluster you want to check the installation progress for by running the following command:

      $ export KUBECONFIG=<cluster_kubeconfig_absolute_path>

    2. Check all the subscriptions present on the single-node cluster and look for the one in the policy you are trying to install through the ClusterGroupUpgrade CR by running the following command:

      $ oc get subs -A | grep -i <subscription_name>
      Example output for cluster-logging policy
      NAMESPACE                              NAME                         PACKAGE                      SOURCE             CHANNEL
openshift-logging                      cluster-logging              cluster-logging              redhat-operators   stable

  3. If one of the managed policies includes a ClusterVersion CR, check the status of platform updates in the current batch by running the following command against the spoke cluster:

    $ oc get clusterversion
    Example output
    NAME      VERSION   AVAILABLE   PROGRESSING   SINCE   STATUS
version   4.4.14.5     True        True          43s     Working towards 4.4.14.7: 71 of 735 done (9% complete)

  4. Check the Operator subscription by running the following command:

    $ oc get subs -n <operator-namespace> <operator-subscription> -ojsonpath="{.status}"

  5. Check the install plans present on the single-node cluster that is associated with the desired subscription by running the following command:

    $ oc get installplan -n <subscription_namespace>
    Example output for cluster-logging Operator
    NAMESPACE                              NAME            CSV                                 APPROVAL   APPROVED
openshift-logging                      install-6khtw   cluster-logging.5.3.3-4             Manual     true (1)
    1 The install plans have their Approval field set to Manual and their Approved field changes from false to true after TALM approves the install plan.

    When TALM is remediating a policy containing a subscription, it automatically approves any install plans attached to that subscription. Where multiple install plans are needed to get the operator to the latest known version, TALM might approve multiple install plans, upgrading through one or more intermediate versions to get to the final version.

  6. Check if the cluster service version for the Operator of the policy that the ClusterGroupUpgrade is installing reached the Succeeded phase by running the following command:

    $ oc get csv -n <operator_namespace>
    Example output for OpenShift Logging Operator
    NAME                    DISPLAY                     VERSION   REPLACES   PHASE
cluster-logging.5.4.2   Red Hat OpenShift Logging   5.4.2                Succeeded

Creating a backup of cluster resources before upgrade

For single-node OpenShift, the Topology Aware Lifecycle Manager (TALM) can create a backup of a deployment before an upgrade. If the upgrade fails, you can recover the previous version and restore a cluster to a working state without requiring a reprovision of applications.

To use the backup feature you first create a ClusterGroupUpgrade CR with the backup field set to true. To ensure that the contents of the backup are up to date, the backup is not taken until you set the enable field in the ClusterGroupUpgrade CR to true.

TALM uses the BackupSucceeded condition to report the status and reasons as follows:

  • true

    Backup is completed for all clusters or the backup run has completed but failed for one or more clusters. If backup fails for any cluster, the update does not proceed for that cluster.

  • false

    Backup is still in progress for one or more clusters or has failed for all clusters. The backup process running in the spoke clusters can have the following statuses:

    • PreparingToStart

      The first reconciliation pass is in progress. The TALM deletes any spoke backup namespace and hub view resources that have been created in a failed upgrade attempt.

    • Starting

      The backup prerequisites and backup job are being created.

    • Active

      The backup is in progress.

    • Succeeded

      The backup succeeded.

    • BackupTimeout

      Artifact backup is partially done.

    • UnrecoverableError

      The backup has ended with a non-zero exit code.

If the backup of a cluster fails and enters the BackupTimeout or UnrecoverableError state, the cluster update does not proceed for that cluster. Updates to other clusters are not affected and continue.

Creating a ClusterGroupUpgrade CR with backup

You can create a backup of a deployment before an upgrade on single-node OpenShift clusters. If the upgrade fails you can use the upgrade-recovery.sh script generated by Topology Aware Lifecycle Manager (TALM) to return the system to its preupgrade state. The backup consists of the following items:

Cluster backup

A snapshot of etcd and static pod manifests.

Content backup

Backups of folders, for example, /etc, /usr/local, /var/lib/kubelet.

Changed files backup

Any files managed by machine-config that have been changed.

Deployment

A pinned ostree deployment.

Images (Optional)

Any container images that are in use.

Prerequisites

  • Install the Topology Aware Lifecycle Manager (TALM).

  • Provision one or more managed clusters.

  • Log in as a user with cluster-admin privileges.

  • Install Red Hat Advanced Cluster Management (RHACM).

It is highly recommended that you create a recovery partition. The following is an example SiteConfig custom resource (CR) for a recovery partition of 50 GB:

nodes:
    - hostName: "node-1.example.com"
    role: "master"
    rootDeviceHints:
        hctl: "0:2:0:0"
        deviceName: /dev/disk/by-id/scsi-3600508b400105e210000900000490000
...
    #Disk /dev/disk/by-id/scsi-3600508b400105e210000900000490000:
    #893.3 GiB, 959119884288 bytes, 1873281024 sectors
    diskPartition:
        - device: /dev/disk/by-id/scsi-3600508b400105e210000900000490000
        partitions:
        - mount_point: /var/recovery
            size: 51200
            start: 800000
Procedure

  1. Save the contents of the ClusterGroupUpgrade CR with the backup and enable fields set to true in the clustergroupupgrades-group-du.yaml file:

    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: du-upgrade-4918
  namespace: ztp-group-du-sno
spec:
  preCaching: true
  backup: true
  clusters:
  - cnfdb1
  - cnfdb2
  enable: true
  managedPolicies:
  - du-upgrade-platform-upgrade
  remediationStrategy:
    maxConcurrency: 2
    timeout: 240

  2. To start the update, apply the ClusterGroupUpgrade CR by running the following command:

    $ oc apply -f clustergroupupgrades-group-du.yaml
Verification

  • Check the status of the upgrade in the hub cluster by running the following command:

    $ oc get cgu -n ztp-group-du-sno du-upgrade-4918 -o jsonpath='{.status}'
    Example output
    {
    "backup": {
        "clusters": [
            "cnfdb2",
            "cnfdb1"
    ],
    "status": {
        "cnfdb1": "Succeeded",
        "cnfdb2": "Failed" (1)
    }
},
"computedMaxConcurrency": 1,
"conditions": [
    {
        "lastTransitionTime": "2022-04-05T10:37:19Z",
        "message": "Backup failed for 1 cluster", (2)
        "reason": "PartiallyDone", (3)
        "status": "True", (4)
        "type": "Succeeded"
    }
],
"precaching": {
    "spec": {}
},
"status": {}
    1 Backup has failed for one cluster.
    2 The message confirms that the backup failed for one cluster.
    3 The backup was partially successful.
    4 The backup process has finished.

Recovering a cluster after a failed upgrade

If an upgrade of a cluster fails, you can manually log in to the cluster and use the backup to return the cluster to its preupgrade state. There are two stages:

Rollback

If the attempted upgrade included a change to the platform OS deployment, you must roll back to the previous version before running the recovery script.

A rollback is only applicable to upgrades from TALM and single-node OpenShift. This process does not apply to rollbacks from any other upgrade type.
Recovery

The recovery shuts down containers and uses files from the backup partition to relaunch containers and restore clusters.

Prerequisites

  • Install the Topology Aware Lifecycle Manager (TALM).

  • Provision one or more managed clusters.

  • Install Red Hat Advanced Cluster Management (RHACM).

  • Log in as a user with cluster-admin privileges.

  • Run an upgrade that is configured for backup.

Procedure

  1. Delete the previously created ClusterGroupUpgrade custom resource (CR) by running the following command:

    $ oc delete cgu/du-upgrade-4918 -n ztp-group-du-sno

  2. Log in to the cluster that you want to recover.

  3. Check the status of the platform OS deployment by running the following command:

    $ ostree admin status
    Example outputs
    [root@lab-test-spoke2-node-0 core]# ostree admin status
* rhcos c038a8f08458bbed83a77ece033ad3c55597e3f64edad66ea12fda18cbdceaf9.0
    Version: 49.84.202202230006-0
    Pinned: yes (1)
    origin refspec: c038a8f08458bbed83a77ece033ad3c55597e3f64edad66ea12fda18cbdceaf9
    1 The current deployment is pinned. A platform OS deployment rollback is not necessary. 
    [root@lab-test-spoke2-node-0 core]# ostree admin status
* rhcos f750ff26f2d5550930ccbe17af61af47daafc8018cd9944f2a3a6269af26b0fa.0
    Version: 410.84.202204050541-0
    origin refspec: f750ff26f2d5550930ccbe17af61af47daafc8018cd9944f2a3a6269af26b0fa
rhcos ad8f159f9dc4ea7e773fd9604c9a16be0fe9b266ae800ac8470f63abc39b52ca.0 (rollback) (1)
    Version: 410.84.202203290245-0
    Pinned: yes (2)
    origin refspec: ad8f159f9dc4ea7e773fd9604c9a16be0fe9b266ae800ac8470f63abc39b52ca
    1 This platform OS deployment is marked for rollback.
    2 The previous deployment is pinned and can be rolled back.

  4. To trigger a rollback of the platform OS deployment, run the following command:

    $ rpm-ostree rollback -r

  5. The first phase of the recovery shuts down containers and restores files from the backup partition to the targeted directories. To begin the recovery, run the following command:

    $ /var/recovery/upgrade-recovery.sh

  6. When prompted, reboot the cluster by running the following command:

    $ systemctl reboot

  7. After the reboot, restart the recovery by running the following command:

    $ /var/recovery/upgrade-recovery.sh  --resume

If the recovery utility fails, you can retry with the --restart option:

$ /var/recovery/upgrade-recovery.sh --restart
Verification

  • To check the status of the recovery run the following command:

    $ oc get clusterversion,nodes,clusteroperator
    Example output
    NAME                                         VERSION   AVAILABLE   PROGRESSING   SINCE   STATUS
clusterversion.config.openshift.io/version   4.4.14.23    True        False         86d     Cluster version is 4.4.14.23 (1)


NAME                          STATUS   ROLES           AGE   VERSION
node/lab-test-spoke1-node-0   Ready    master,worker   86d   v1.22.3+b93fd35 (2)

NAME                                                                           VERSION   AVAILABLE   PROGRESSING   DEGRADED   SINCE   MESSAGE
clusteroperator.config.openshift.io/authentication                             4.4.14.23    True        False         False      2d7h    (3)
clusteroperator.config.openshift.io/baremetal                                  4.4.14.23    True        False         False      86d


..............
    1 The cluster version is available and has the correct version.
    2 The node status is Ready.
    3 The ClusterOperator object’s availability is True.

Using the container image pre-cache feature

Single-node OpenShift clusters might have limited bandwidth to access the container image registry, which can cause a timeout before the updates are completed.

The time of the update is not set by TALM. You can apply the ClusterGroupUpgrade CR at the beginning of the update by manual application or by external automation.

The container image pre-caching starts when the preCaching field is set to true in the ClusterGroupUpgrade CR.

TALM uses the PrecacheSpecValid condition to report status information as follows:

  • true

    The pre-caching spec is valid and consistent.

  • false

    The pre-caching spec is incomplete.

TALM uses the PrecachingSucceeded condition to report status information as follows:

  • true

    TALM has concluded the pre-caching process. If pre-caching fails for any cluster, the update fails for that cluster but proceeds for all other clusters. A message informs you if pre-caching has failed for any clusters.

  • false

    Pre-caching is still in progress for one or more clusters or has failed for all clusters.

After a successful pre-caching process, you can start remediating policies. The remediation actions start when the enable field is set to true. If there is a pre-caching failure on a cluster, the upgrade fails for that cluster. The upgrade process continues for all other clusters that have a successful pre-cache.

The pre-caching process can be in the following statuses:

  • NotStarted

    This is the initial state all clusters are automatically assigned to on the first reconciliation pass of the ClusterGroupUpgrade CR. In this state, TALM deletes any pre-caching namespace and hub view resources of spoke clusters that remain from previous incomplete updates. TALM then creates a new ManagedClusterView resource for the spoke pre-caching namespace to verify its deletion in the PrecachePreparing state.

  • PreparingToStart

    Cleaning up any remaining resources from previous incomplete updates is in progress.

  • Starting

    Pre-caching job prerequisites and the job are created.

  • Active

    The job is in "Active" state.

  • Succeeded

    The pre-cache job succeeded.

  • PrecacheTimeout

    The artifact pre-caching is partially done.

  • UnrecoverableError

    The job ends with a non-zero exit code.

Using the container image pre-cache filter

The pre-cache feature typically downloads more images than a cluster needs for an update. You can control which pre-cache images are downloaded to a cluster. This decreases download time, and saves bandwidth and storage.

You can see a list of all images to be downloaded using the following command:

$ oc adm release info <ocp-version>

The following ConfigMap example shows how you can exclude images using the excludePrecachePatterns field.

apiVersion: v1
kind: ConfigMap
metadata:
  name: cluster-group-upgrade-overrides
data:
  excludePrecachePatterns: |
    azure (1)
    aws
    vsphere
    alibaba
1 TALM excludes all images with names that include any of the patterns listed here.

Creating a ClusterGroupUpgrade CR with pre-caching

For single-node OpenShift, the pre-cache feature allows the required container images to be present on the spoke cluster before the update starts.

For pre-caching, TALM uses the spec.remediationStrategy.timeout value from the ClusterGroupUpgrade CR. You must set a timeout value that allows sufficient time for the pre-caching job to complete. When you enable the ClusterGroupUpgrade CR after pre-caching has completed, you can change the timeout value to a duration that is appropriate for the update.
Prerequisites

  • Install the Topology Aware Lifecycle Manager (TALM).

  • Provision one or more managed clusters.

  • Log in as a user with cluster-admin privileges.

Procedure

  1. Save the contents of the ClusterGroupUpgrade CR with the preCaching field set to true in the clustergroupupgrades-group-du.yaml file:

    apiVersion: ran.openshift.io/v1alpha1
kind: ClusterGroupUpgrade
metadata:
  name: du-upgrade-4918
  namespace: ztp-group-du-sno
spec:
  preCaching: true (1)
  clusters:
  - cnfdb1
  - cnfdb2
  enable: false
  managedPolicies:
  - du-upgrade-platform-upgrade
  remediationStrategy:
    maxConcurrency: 2
    timeout: 240
    1 The preCaching field is set to true, which enables TALM to pull the container images before starting the update.

  2. When you want to start pre-caching, apply the ClusterGroupUpgrade CR by running the following command:

    $ oc apply -f clustergroupupgrades-group-du.yaml
Verification

  1. Check if the ClusterGroupUpgrade CR exists in the hub cluster by running the following command:

    $ oc get cgu -A
    Example output
    NAMESPACE          NAME              AGE   STATE        DETAILS
ztp-group-du-sno   du-upgrade-4918   10s   InProgress   Precaching is required and not done (1)
    1 The CR is created.

  2. Check the status of the pre-caching task by running the following command:

    $ oc get cgu -n ztp-group-du-sno du-upgrade-4918 -o jsonpath='{.status}'
    Example output
    {
  "conditions": [
    {
      "lastTransitionTime": "2022-01-27T19:07:24Z",
      "message": "Precaching is required and not done",
      "reason": "InProgress",
      "status": "False",
      "type": "PrecachingSucceeded"
    },
    {
      "lastTransitionTime": "2022-01-27T19:07:34Z",
      "message": "Pre-caching spec is valid and consistent",
      "reason": "PrecacheSpecIsWellFormed",
      "status": "True",
      "type": "PrecacheSpecValid"
    }
  ],
  "precaching": {
    "clusters": [
      "cnfdb1" (1)
      "cnfdb2"
    ],
    "spec": {
      "platformImage": "image.example.io"},
    "status": {
      "cnfdb1": "Active"
      "cnfdb2": "Succeeded"}
    }
}
    1 Displays the list of identified clusters.

  3. Check the status of the pre-caching job by running the following command on the spoke cluster:

    $ oc get jobs,pods -n openshift-talo-pre-cache
    Example output
    NAME                  COMPLETIONS   DURATION   AGE
job.batch/pre-cache   0/1           3m10s      3m10s

NAME                     READY   STATUS    RESTARTS   AGE
pod/pre-cache--1-9bmlr   1/1     Running   0          3m10s

  4. Check the status of the ClusterGroupUpgrade CR by running the following command:

    $ oc get cgu -n ztp-group-du-sno du-upgrade-4918 -o jsonpath='{.status}'
    Example output
    "conditions": [
    {
      "lastTransitionTime": "2022-01-27T19:30:41Z",
      "message": "The ClusterGroupUpgrade CR has all clusters compliant with all the managed policies",
      "reason": "UpgradeCompleted",
      "status": "True",
      "type": "Ready"
    },
    {
      "lastTransitionTime": "2022-01-27T19:28:57Z",
      "message": "Precaching is completed",
      "reason": "PrecachingCompleted",
      "status": "True",
      "type": "PrecachingSucceeded" (1)
    }
    1 The pre-cache tasks are done.

Troubleshooting the Topology Aware Lifecycle Manager

The Topology Aware Lifecycle Manager (TALM) is an OKD Operator that remediates RHACM policies. When issues occur, use the oc adm must-gather command to gather details and logs and to take steps in debugging the issues.

For more information about related topics, see the following documentation:

General troubleshooting

You can determine the cause of the problem by reviewing the following questions:

To ensure that the ClusterGroupUpgrade configuration is functional, you can do the following:

  1. Create the ClusterGroupUpgrade CR with the spec.enable field set to false.

  2. Wait for the status to be updated and go through the troubleshooting questions.

  3. If everything looks as expected, set the spec.enable field to true in the ClusterGroupUpgrade CR.

After you set the spec.enable field to true in the ClusterUpgradeGroup CR, the update procedure starts and you cannot edit the CR’s spec fields anymore.

Cannot modify the ClusterUpgradeGroup CR

Issue

You cannot edit the ClusterUpgradeGroup CR after enabling the update.

Resolution

Restart the procedure by performing the following steps:

  1. Remove the old ClusterGroupUpgrade CR by running the following command:

    $ oc delete cgu -n <ClusterGroupUpgradeCR_namespace> <ClusterGroupUpgradeCR_name>

  2. Check and fix the existing issues with the managed clusters and policies.

    1. Ensure that all the clusters are managed clusters and available.

    2. Ensure that all the policies exist and have the spec.remediationAction field set to inform.

  3. Create a new ClusterGroupUpgrade CR with the correct configurations.

    $ oc apply -f <ClusterGroupUpgradeCR_YAML>

Managed policies

Checking managed policies on the system

Issue

You want to check if you have the correct managed policies on the system.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -ojsonpath='{.spec.managedPolicies}'
Example output
["group-du-sno-validator-du-validator-policy", "policy2-common-nto-sub-policy", "policy3-common-ptp-sub-policy"]

Checking remediationAction mode

Issue

You want to check if the remediationAction field is set to inform in the spec of the managed policies.

Resolution

Run the following command:

$ oc get policies --all-namespaces
Example output
NAMESPACE   NAME                                                 REMEDIATION ACTION   COMPLIANCE STATE   AGE
default     policy1-common-cluster-version-policy                inform               NonCompliant       5d21h
default     policy2-common-nto-sub-policy                        inform               Compliant          5d21h
default     policy3-common-ptp-sub-policy                        inform               NonCompliant       5d21h
default     policy4-common-sriov-sub-policy                      inform               NonCompliant       5d21h

Checking policy compliance state

Issue

You want to check the compliance state of policies.

Resolution

Run the following command:

$ oc get policies --all-namespaces
Example output
NAMESPACE   NAME                                                 REMEDIATION ACTION   COMPLIANCE STATE   AGE
default     policy1-common-cluster-version-policy                inform               NonCompliant       5d21h
default     policy2-common-nto-sub-policy                        inform               Compliant          5d21h
default     policy3-common-ptp-sub-policy                        inform               NonCompliant       5d21h
default     policy4-common-sriov-sub-policy                      inform               NonCompliant       5d21h

Clusters

Checking if managed clusters are present

Issue

You want to check if the clusters in the ClusterGroupUpgrade CR are managed clusters.

Resolution

Run the following command:

$ oc get managedclusters
Example output
NAME            HUB ACCEPTED   MANAGED CLUSTER URLS                    JOINED   AVAILABLE   AGE
local-cluster   true           https://api.hub.example.com:6443        True     Unknown     13d
spoke1          true           https://api.spoke1.example.com:6443     True     True        13d
spoke3          true           https://api.spoke3.example.com:6443     True     True        27h

  1. Alternatively, check the TALM manager logs:

    1. Get the name of the TALM manager by running the following command:

      $ oc get pod -n openshift-operators
      Example output
      NAME                                                         READY   STATUS    RESTARTS   AGE
cluster-group-upgrades-controller-manager-75bcc7484d-8k8xp   2/2     Running   0          45m

    2. Check the TALM manager logs by running the following command:

      $ oc logs -n openshift-operators \
cluster-group-upgrades-controller-manager-75bcc7484d-8k8xp -c manager
      Example output
      ERROR	controller-runtime.manager.controller.clustergroupupgrade	Reconciler error	{"reconciler group": "ran.openshift.io", "reconciler kind": "ClusterGroupUpgrade", "name": "lab-upgrade", "namespace": "default", "error": "Cluster spoke5555 is not a ManagedCluster"} (1)
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).processNextWorkItem
      1 The error message shows that the cluster is not a managed cluster.

Checking if managed clusters are available

Issue

You want to check if the managed clusters specified in the ClusterGroupUpgrade CR are available.

Resolution

Run the following command:

$ oc get managedclusters
Example output
NAME            HUB ACCEPTED   MANAGED CLUSTER URLS                    JOINED   AVAILABLE   AGE
local-cluster   true           https://api.hub.testlab.com:6443        True     Unknown     13d
spoke1          true           https://api.spoke1.testlab.com:6443     True     True        13d (1)
spoke3          true           https://api.spoke3.testlab.com:6443     True     True        27h (1)
1 The value of the AVAILABLE field is True for the managed clusters.

Checking clusterLabelSelector

Issue

You want to check if the clusterLabelSelector field specified in the ClusterGroupUpgrade CR matches at least one of the managed clusters.

Resolution

Run the following command:

$ oc get managedcluster --selector=upgrade=true (1)
1 The label for the clusters you want to update is upgrade:true.
Example output
NAME            HUB ACCEPTED   MANAGED CLUSTER URLS                     JOINED    AVAILABLE   AGE
spoke1          true           https://api.spoke1.testlab.com:6443      True     True        13d
spoke3          true           https://api.spoke3.testlab.com:6443      True     True        27h

Checking if canary clusters are present

Issue

You want to check if the canary clusters are present in the list of clusters.

Example ClusterGroupUpgrade CR
spec:
    remediationStrategy:
        canaries:
        - spoke3
        maxConcurrency: 2
        timeout: 240
    clusterLabelSelectors:
      - matchLabels:
          upgrade: true
Resolution

Run the following commands:

$ oc get cgu lab-upgrade -ojsonpath='{.spec.clusters}'
Example output
["spoke1", "spoke3"]

  1. Check if the canary clusters are present in the list of clusters that match clusterLabelSelector labels by running the following command:

    $ oc get managedcluster --selector=upgrade=true
    Example output
    NAME            HUB ACCEPTED   MANAGED CLUSTER URLS   JOINED    AVAILABLE   AGE
spoke1          true           https://api.spoke1.testlab.com:6443   True     True        13d
spoke3          true           https://api.spoke3.testlab.com:6443   True     True        27h

A cluster can be present in spec.clusters and also be matched by the spec.clusterLabelSelector label.

Checking the pre-caching status on spoke clusters

  1. Check the status of pre-caching by running the following command on the spoke cluster:

    $ oc get jobs,pods -n openshift-talo-pre-cache

Remediation Strategy

Checking if remediationStrategy is present in the ClusterGroupUpgrade CR

Issue

You want to check if the remediationStrategy is present in the ClusterGroupUpgrade CR.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -ojsonpath='{.spec.remediationStrategy}'
Example output
{"maxConcurrency":2, "timeout":240}

Checking if maxConcurrency is specified in the ClusterGroupUpgrade CR

Issue

You want to check if the maxConcurrency is specified in the ClusterGroupUpgrade CR.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -ojsonpath='{.spec.remediationStrategy.maxConcurrency}'
Example output
2

Topology Aware Lifecycle Manager

Checking condition message and status in the ClusterGroupUpgrade CR

Issue

You want to check the value of the status.conditions field in the ClusterGroupUpgrade CR.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -ojsonpath='{.status.conditions}'
Example output
{"lastTransitionTime":"2022-02-17T22:25:28Z", "message":"Missing managed policies:[policyList]", "reason":"NotAllManagedPoliciesExist", "status":"False", "type":"Validated"}

Checking corresponding copied policies

Issue

You want to check if every policy from status.managedPoliciesForUpgrade has a corresponding policy in status.copiedPolicies.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -oyaml
Example output
status:
  …
  copiedPolicies:
  - lab-upgrade-policy3-common-ptp-sub-policy
  managedPoliciesForUpgrade:
  - name: policy3-common-ptp-sub-policy
    namespace: default

Checking if status.remediationPlan was computed

Issue

You want to check if status.remediationPlan is computed.

Resolution

Run the following command:

$ oc get cgu lab-upgrade -ojsonpath='{.status.remediationPlan}'
Example output
[["spoke2", "spoke3"]]

Errors in the TALM manager container

Issue

You want to check the logs of the manager container of TALM.

Resolution

Run the following command:

$ oc logs -n openshift-operators \
cluster-group-upgrades-controller-manager-75bcc7484d-8k8xp -c manager
Example output
ERROR	controller-runtime.manager.controller.clustergroupupgrade	Reconciler error	{"reconciler group": "ran.openshift.io", "reconciler kind": "ClusterGroupUpgrade", "name": "lab-upgrade", "namespace": "default", "error": "Cluster spoke5555 is not a ManagedCluster"} (1)
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).processNextWorkItem
1 Displays the error.

Clusters are not compliant to some policies after a ClusterGroupUpgrade CR has completed

Issue

The policy compliance status that TALM uses to decide if remediation is needed has not yet fully updated for all clusters. This may be because:

  • The CGU was run too soon after a policy was created or updated.

  • The remediation of a policy affects the compliance of subsequent policies in the ClusterGroupUpgrade CR.

Resolution

Create and apply a new ClusterGroupUpdate CR with the same specification.

Auto-created ClusterGroupUpgrade CR in the GitOps ZTP workflow has no managed policies

Issue

If there are no policies for the managed cluster when the cluster becomes Ready, a ClusterGroupUpgrade CR with no policies is auto-created. Upon completion of the ClusterGroupUpgrade CR, the managed cluster is labeled as ztp-done. If the PolicyGenTemplate CRs were not pushed to the Git repository within the required time after SiteConfig resources were pushed, this might result in no policies being available for the target cluster when the cluster became Ready.

Resolution

Verify that the policies you want to apply are available on the hub cluster, then create a ClusterGroupUpgrade CR with the required policies.

You can either manually create the ClusterGroupUpgrade CR or trigger auto-creation again. To trigger auto-creation of the ClusterGroupUpgrade CR, remove the ztp-done label from the cluster and delete the empty ClusterGroupUpgrade CR that was previously created in the zip-install namespace.

Pre-caching has failed

Issue

Pre-caching might fail for one of the following reasons:

  • There is not enough free space on the node.

  • For a disconnected environment, the pre-cache image has not been properly mirrored.

  • There was an issue when creating the pod.

Resolution

  1. To check if pre-caching has failed due to insufficient space, check the log of the pre-caching pod in the node.

    1. Find the name of the pod using the following command:

      $ oc get pods -n openshift-talo-pre-cache

    2. Check the logs to see if the error is related to insufficient space using the following command:

      $ oc logs -n openshift-talo-pre-cache <pod name>

  2. If there is no log, check the pod status using the following command:

    $ oc describe pod -n openshift-talo-pre-cache <pod name>

  3. If the pod does not exist, check the job status to see why it could not create a pod using the following command:

    $ oc describe job -n openshift-talo-pre-cache pre-cache
Additional resources