Chapter 1. Troubleshooting
Before using the Troubleshooting guide, you can run the oc adm must-gather
command to gather details, logs, and take steps in debugging issues. For more details, see Running the must-gather command to troubleshoot.
Additionally, check your role-based access. See Role-based access control for details.
1.1. Documented troubleshooting
View the list of troubleshooting topics for Red Hat Advanced Cluster Management for Kubernetes:
Installation
To view the main documentation for the installing tasks, see Installing.
Backup and restore
To view the main documentation for backup and restore, see Backup and restore.
Cluster management
To view the main documentation about managing your clusters, see Cluster lifecycle architecture or The multicluster engine operator cluster lifecycle overview.
- Troubleshooting an offline cluster
- Troubleshooting a managed cluster import failure
- Troubleshooting cluster with pending import status
- Troubleshooting imported clusters offline after certificate change
- Troubleshooting cluster status changing from offline to available
- Troubleshooting cluster creation on VMware vSphere
- Troubleshooting cluster in console with pending or failed status
- Troubleshooting OpenShift Container Platform version 3.11 cluster import failure
- Troubleshooting Klusterlet with degraded conditions
- Troubleshooting Klusterlet application manager on managed clusters
- Troubleshooting Object storage channel secret
- Namespace remains after deleting a cluster
- Auto-import-secret-exists error when importing a cluster
Application management
To view the main documentation about application management, see Managing applications.
Governance
To view the security guide, see Risk and compliance.
Console observability
Console observability includes Search, along with header and navigation function. To view the observability guide, see Observability in the console.
Health metrics
Submariner networking and service discovery
This section lists the Submariner troubleshooting procedures that can occur when using Submariner with Red Hat Advanced Cluster Management or multicluster engine operator. For general Submariner troubleshooting information, see Troubleshooting in the Submariner documentation.
To view the main documentation for the Submariner networking service and service discovery, see Submariner multicluster networking and service discovery.
1.2. Running the must-gather command to troubleshoot
To get started with troubleshooting, learn about the troubleshooting scenarios for users to run the must-gather
command to debug the issues, then see the procedures to start using the command.
Required access: Cluster administrator
1.2.1. Must-gather scenarios
Scenario one: Use the Documented troubleshooting section to see if a solution to your problem is documented. The guide is organized by the major functions of the product.
With this scenario, you check the guide to see if your solution is in the documentation. For instance, for trouble with creating a cluster, you might find a solution in the Manage cluster section.
-
Scenario two: If your problem is not documented with steps to resolve, run the
must-gather
command and use the output to debug the issue. -
Scenario three: If you cannot debug the issue using your output from the
must-gather
command, then share your output with Red Hat Support.
1.2.2. Must-gather procedure
See the following procedure to start using the must-gather
command:
-
Learn about the
must-gather
command and install the prerequisites that you need at Gathering data about your cluster in the Red Hat OpenShift Container Platform documentation. Log in to your cluster. Add the Red Hat Advanced Cluster Management for Kubernetes image that is used for gathering data and the directory. Run the following command, where you insert the image and the directory for the output:
oc adm must-gather --image=registry.redhat.io/rhacm2/acm-must-gather-rhel8:v2.7 --dest-dir=<directory>
For the usual use-case, you should run the
must-gather
while you are logged into your hub cluster.Note: If you want to check your managed clusters, find the
gather-managed.log
file that is located in thecluster-scoped-resources
directory:<your-directory>/cluster-scoped-resources/gather-managed.log>
Check for managed clusters that are not set
True
for the JOINED and AVAILABLE column. You can run themust-gather
command on those clusters that are not connected withTrue
status.Go to your specified directory to see your output, which is organized in the following levels:
-
Two peer levels:
cluster-scoped-resources
andnamespace
resources. - Sub-level for each: API group for the custom resource definitions for both cluster-scope and namespace-scoped resources.
-
Next level for each: YAML file sorted by
kind
.
-
Two peer levels:
1.2.3. Must-gather in a disconnected environment
Complete the following steps to run the must-gather
command in a disconnected environment:
- In a disconnected environment, mirror the Red Hat operator catalog images into their mirror registry. For more information, see Install on disconnected networks.
Run the following command to extract logs, which reference the image from their mirror registry:
REGISTRY=registry.example.com:5000 IMAGE=$REGISTRY/rhacm2/acm-must-gather-rhel8@sha256:ff9f37eb400dc1f7d07a9b6f2da9064992934b69847d17f59e385783c071b9d8 oc adm must-gather --image=$IMAGE --dest-dir=./data
Run the following commands to collect all of the information:
REGISTRY=<internal.repo.address:port> IMAGE1=$REGISTRY/rhacm2/acm-must-gather-rhel8:v2.7 IMAGE2=$REGISTRY/multicluster-engine/must-gather-rhel8:v2.2 oc adm must-gather --image=$IMAGE1 --dest-dir=<directory> oc adm must-gather --image=$IMAGE2 --dest-dir=<directory>
If you experience issues with one of the currently supported releases, or the product documentation, go to Red Hat Support where you can troubleshoot further, view Knowledgebase articles, connect with the Support Team, or open a case. You must log in with your Red Hat credentials.
1.3. Troubleshooting installation status stuck in installing or pending
When installing Red Hat Advanced Cluster Management, the MultiClusterHub
remains in Installing
phase, or multiple pods maintain a Pending
status.
1.3.1. Symptom: Stuck in Pending status
More than ten minutes passed since you installed MultiClusterHub
and one or more components from the status.components
field of the MultiClusterHub
resource report ProgressDeadlineExceeded
. Resource constraints on the cluster might be the issue.
Check the pods in the namespace where Multiclusterhub
was installed. You might see Pending
with a status similar to the following:
reason: Unschedulable message: '0/6 nodes are available: 3 Insufficient cpu, 3 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate.'
In this case, the worker nodes resources are not sufficient in the cluster to run the product.
1.3.2. Resolving the problem: Adjust worker node sizing
If you have this problem, then your cluster needs to be updated with either larger or more worker nodes. See Sizing your cluster for guidelines on sizing your cluster.
1.4. Troubleshooting reinstallation failure
When reinstalling Red Hat Advanced Cluster Management for Kubernetes, the pods do not start.
1.4.1. Symptom: Reinstallation failure
If your pods do not start after you install Red Hat Advanced Cluster Management, it is likely that Red Hat Advanced Cluster Management was previously installed, and not all of the pieces were removed before you attempted this installation.
In this case, the pods do not start after completing the installation process.
1.4.2. Resolving the problem: Reinstallation failure
If you have this problem, complete the following steps:
- Run the uninstallation process to remove the current components by following the steps in Uninstalling.
- Install the Helm CLI binary version 3.2.0, or later, by following the instructions at Installing Helm.
-
Ensure that your Red Hat OpenShift Container Platform CLI is configured to run
oc
commands. See Getting started with the OpenShift CLI in the OpenShift Container Platform documentation for more information about how to configure theoc
commands. Copy the following script into a file:
#!/bin/bash ACM_NAMESPACE=<namespace> oc delete mch --all -n $ACM_NAMESPACE helm ls --namespace $ACM_NAMESPACE | cut -f 1 | tail -n +2 | xargs -n 1 helm delete --namespace $ACM_NAMESPACE oc delete apiservice v1beta1.webhook.certmanager.k8s.io v1.admission.cluster.open-cluster-management.io v1.admission.work.open-cluster-management.io oc delete clusterimageset --all oc delete configmap -n $ACM_NAMESPACE cert-manager-controller cert-manager-cainjector-leader-election cert-manager-cainjector-leader-election-core oc delete consolelink acm-console-link oc delete crd klusterletaddonconfigs.agent.open-cluster-management.io placementbindings.policy.open-cluster-management.io policies.policy.open-cluster-management.io userpreferences.console.open-cluster-management.io searchservices.search.acm.com discoveredclusters.discovery.open-cluster-management.io discoveryconfigs.discovery.open-cluster-management.io oc delete mutatingwebhookconfiguration cert-manager-webhook cert-manager-webhook-v1alpha1 ocm-mutating-webhook managedclustermutators.admission.cluster.open-cluster-management.io oc delete oauthclient multicloudingress oc delete rolebinding -n kube-system cert-manager-webhook-webhook-authentication-reader oc delete scc kui-proxy-scc oc delete validatingwebhookconfiguration cert-manager-webhook cert-manager-webhook-v1alpha1 channels.apps.open.cluster.management.webhook.validator application-webhook-validator multiclusterhub-operator-validating-webhook ocm-validating-webhook
Replace
<namespace>
in the script with the name of the namespace where Red Hat Advanced Cluster Management was installed. Ensure that you specify the correct namespace, as the namespace is cleaned out and deleted.- Run the script to remove the artifacts from the previous installation.
- Run the installation. See Installing while connected online.
1.5. Troubleshooting an offline cluster
There are a few common causes for a cluster showing an offline status.
1.5.1. Symptom: Cluster status is offline
After you complete the procedure for creating a cluster, you cannot access it from the Red Hat Advanced Cluster Management console, and it shows a status of offline
.
1.5.2. Resolving the problem: Cluster status is offline
Determine if the managed cluster is available. You can check this in the Clusters area of the Red Hat Advanced Cluster Management console.
If it is not available, try restarting the managed cluster.
If the managed cluster status is still offline, complete the following steps:
-
Run the
oc get managedcluster <cluster_name> -o yaml
command on the hub cluster. Replace<cluster_name>
with the name of your cluster. -
Find the
status.conditions
section. -
Check the messages for
type: ManagedClusterConditionAvailable
and resolve any problems.
-
Run the
1.6. Troubleshooting a managed cluster import failure
If your cluster import fails, there are a few steps that you can take to determine why the cluster import failed.
1.6.1. Symptom: Imported cluster not available
After you complete the procedure for importing a cluster, you cannot access it from the Red Hat Advanced Cluster Management for Kubernetes console.
1.6.2. Resolving the problem: Imported cluster not available
There can be a few reasons why an imported cluster is not available after an attempt to import it. If the cluster import fails, complete the following steps, until you find the reason for the failed import:
On the Red Hat Advanced Cluster Management hub cluster, run the following command to ensure that the Red Hat Advanced Cluster Management import controller is running.
kubectl -n multicluster-engine get pods -l app=managedcluster-import-controller-v2
You should see two pods that are running. If either of the pods is not running, run the following command to view the log to determine the reason:
kubectl -n multicluster-engine logs -l app=managedcluster-import-controller-v2 --tail=-1
On the Red Hat Advanced Cluster Management hub cluster, run the following command to determine if the managed cluster import secret was generated successfully by the Red Hat Advanced Cluster Management import controller:
kubectl -n <managed_cluster_name> get secrets <managed_cluster_name>-import
If the import secret does not exist, run the following command to view the log entries for the import controller and determine why it was not created:
kubectl -n multicluster-engine logs -l app=managedcluster-import-controller-v2 --tail=-1 | grep importconfig-controller
On the Red Hat Advanced Cluster Management hub cluster, if your managed cluster is
local-cluster
, provisioned by Hive, or has an auto-import secret, run the following command to check the import status of the managed cluster.kubectl get managedcluster <managed_cluster_name> -o=jsonpath='{range .status.conditions[*]}{.type}{"\t"}{.status}{"\t"}{.message}{"\n"}{end}' | grep ManagedClusterImportSucceeded
If the condition
ManagedClusterImportSucceeded
is nottrue
, the result of the command indicates the reason for the failure.- Check the Klusterlet status of the managed cluster for a degraded condition. See Troubleshooting Klusterlet with degraded conditions to find the reason that the Klusterlet is degraded.
1.7. Troubleshooting cluster with pending import status
If you receive Pending import continually on the console of your cluster, follow the procedure to troubleshoot the problem.
1.7.1. Symptom: Cluster with pending import status
After importing a cluster by using the Red Hat Advanced Cluster Management console, the cluster appears in the console with a status of Pending import.
1.7.2. Identifying the problem: Cluster with pending import status
Run the following command on the managed cluster to view the Kubernetes pod names that are having the issue:
kubectl get pod -n open-cluster-management-agent | grep klusterlet-registration-agent
Run the following command on the managed cluster to find the log entry for the error:
kubectl logs <registration_agent_pod> -n open-cluster-management-agent
Replace registration_agent_pod with the pod name that you identified in step 1.
-
Search the returned results for text that indicates there was a networking connectivity problem. Example includes:
no such host
.
1.7.3. Resolving the problem: Cluster with pending import status
Retrieve the port number that is having the problem by entering the following command on the hub cluster:
oc get infrastructure cluster -o yaml | grep apiServerURL
Ensure that the hostname from the managed cluster can be resolved, and that outbound connectivity to the host and port is occurring.
If the communication cannot be established by the managed cluster, the cluster import is not complete. The cluster status for the managed cluster is Pending import.
1.8. Troubleshooting cluster with already exists error
If you are unable to import an OpenShift Container Platform cluster into Red Hat Advanced Cluster Management MultiClusterHub
and receive an AlreadyExists
error, follow the procedure to troubleshoot the problem.
1.8.1. Symptom: Already exists error log when importing OpenShift Container Platform cluster
An error log shows up when importing an OpenShift Container Platform cluster into Red Hat Advanced Cluster Management MultiClusterHub
:
error log: Warning: apiextensions.k8s.io/v1beta1 CustomResourceDefinition is deprecated in v1.16+, unavailable in v1.22+; use apiextensions.k8s.io/v1 CustomResourceDefinition Error from server (AlreadyExists): error when creating "STDIN": customresourcedefinitions.apiextensions.k8s.io "klusterlets.operator.open-cluster-management.io" already exists The cluster cannot be imported because its Klusterlet CRD already exists. Either the cluster was already imported, or it was not detached completely during a previous detach process. Detach the existing cluster before trying the import again."
1.8.2. Identifying the problem: Already exists when importing OpenShift Container Platform cluster
Check if there are any Red Hat Advanced Cluster Management-related resources on the cluster that you want to import to new the Red Hat Advanced Cluster Management MultiClusterHub
by running the following commands:
oc get all -n open-cluster-management-agent oc get all -n open-cluster-management-agent-addon
1.8.3. Resolving the problem: Already exists when importing OpenShift Container Platform cluster
Run the following commands to remove pre-existing resources:
oc delete namespaces open-cluster-management-agent open-cluster-management-agent-addon --wait=false oc get crds | grep open-cluster-management.io | awk '{print $1}' | xargs oc delete crds --wait=false oc get crds | grep open-cluster-management.io | awk '{print $1}' | xargs oc patch crds --type=merge -p '{"metadata":{"finalizers": []}}'
1.9. Troubleshooting cluster creation on VMware vSphere
If you experience a problem when creating a Red Hat OpenShift Container Platform cluster on VMware vSphere, see the following troubleshooting information to see if one of them addresses your problem.
Note: Sometimes when the cluster creation process fails on VMware vSphere, the link is not enabled for you to view the logs. If this happens, you can identify the problem by viewing the log of the hive-controllers
pod. The hive-controllers
log is in the hive
namespace.
1.9.1. Managed cluster creation fails with certificate IP SAN error
1.9.1.1. Symptom: Managed cluster creation fails with certificate IP SAN error
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails with an error message that indicates a certificate IP SAN error.
1.9.1.2. Identifying the problem: Managed cluster creation fails with certificate IP SAN error
The deployment of the managed cluster fails and returns the following errors in the deployment log:
time="2020-08-07T15:27:55Z" level=error msg="Error: error setting up new vSphere SOAP client: Post https://147.1.1.1/sdk: x509: cannot validate certificate for xx.xx.xx.xx because it doesn't contain any IP SANs" time="2020-08-07T15:27:55Z" level=error
1.9.1.3. Resolving the problem: Managed cluster creation fails with certificate IP SAN error
Use the VMware vCenter server fully-qualified host name instead of the IP address in the credential. You can also update the VMware vCenter CA certificate to contain the IP SAN.
1.9.2. Managed cluster creation fails with unknown certificate authority
1.9.2.1. Symptom: Managed cluster creation fails with unknown certificate authority
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because the certificate is signed by an unknown authority.
1.9.2.2. Identifying the problem: Managed cluster creation fails with unknown certificate authority
The deployment of the managed cluster fails and returns the following errors in the deployment log:
Error: error setting up new vSphere SOAP client: Post https://vspherehost.com/sdk: x509: certificate signed by unknown authority"
1.9.2.3. Resolving the problem: Managed cluster creation fails with unknown certificate authority
Ensure you entered the correct certificate from the certificate authority when creating the credential.
1.9.3. Managed cluster creation fails with expired certificate
1.9.3.1. Symptom: Managed cluster creation fails with expired certificate
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because the certificate is expired or is not yet valid.
1.9.3.2. Identifying the problem: Managed cluster creation fails with expired certificate
The deployment of the managed cluster fails and returns the following errors in the deployment log:
x509: certificate has expired or is not yet valid
1.9.3.3. Resolving the problem: Managed cluster creation fails with expired certificate
Ensure that the time on your ESXi hosts is synchronized.
1.9.4. Managed cluster creation fails with insufficient privilege for tagging
1.9.4.1. Symptom: Managed cluster creation fails with insufficient privilege for tagging
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because there is insufficient privilege to use tagging.
1.9.4.2. Identifying the problem: Managed cluster creation fails with insufficient privilege for tagging
The deployment of the managed cluster fails and returns the following errors in the deployment log:
time="2020-08-07T19:41:58Z" level=debug msg="vsphere_tag_category.category: Creating..." time="2020-08-07T19:41:58Z" level=error time="2020-08-07T19:41:58Z" level=error msg="Error: could not create category: POST https://vspherehost.com/rest/com/vmware/cis/tagging/category: 403 Forbidden" time="2020-08-07T19:41:58Z" level=error time="2020-08-07T19:41:58Z" level=error msg=" on ../tmp/openshift-install-436877649/main.tf line 54, in resource \"vsphere_tag_category\" \"category\":" time="2020-08-07T19:41:58Z" level=error msg=" 54: resource \"vsphere_tag_category\" \"category\" {"
1.9.4.3. Resolving the problem: Managed cluster creation fails with insufficient privilege for tagging
Ensure that your VMware vCenter required account privileges are correct. See Image registry removed during information for more information.
1.9.5. Managed cluster creation fails with invalid dnsVIP
1.9.5.1. Symptom: Managed cluster creation fails with invalid dnsVIP
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because there is an invalid dnsVIP.
1.9.5.2. Identifying the problem: Managed cluster creation fails with invalid dnsVIP
If you see the following message when trying to deploy a new managed cluster with VMware vSphere, it is because you have an older OpenShift Container Platform release image that does not support VMware Installer Provisioned Infrastructure (IPI):
failed to fetch Master Machines: failed to load asset \\\"Install Config\\\": invalid \\\"install-config.yaml\\\" file: platform.vsphere.dnsVIP: Invalid value: \\\"\\\": \\\"\\\" is not a valid IP
1.9.5.3. Resolving the problem: Managed cluster creation fails with invalid dnsVIP
Select a release image from a later version of OpenShift Container Platform that supports VMware Installer Provisioned Infrastructure.
1.9.6. Managed cluster creation fails with incorrect network type
1.9.6.1. Symptom: Managed cluster creation fails with incorrect network type
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because there is an incorrect network type specified.
1.9.6.2. Identifying the problem: Managed cluster creation fails with incorrect network type
If you see the following message when trying to deploy a new managed cluster with VMware vSphere, it is because you have an older OpenShift Container Platform image that does not support VMware Installer Provisioned Infrastructure (IPI):
time="2020-08-11T14:31:38-04:00" level=debug msg="vsphereprivate_import_ova.import: Creating..." time="2020-08-11T14:31:39-04:00" level=error time="2020-08-11T14:31:39-04:00" level=error msg="Error: rpc error: code = Unavailable desc = transport is closing" time="2020-08-11T14:31:39-04:00" level=error time="2020-08-11T14:31:39-04:00" level=error time="2020-08-11T14:31:39-04:00" level=fatal msg="failed to fetch Cluster: failed to generate asset \"Cluster\": failed to create cluster: failed to apply Terraform: failed to complete the change"
1.9.6.3. Resolving the problem: Managed cluster creation fails with incorrect network type
Select a valid VMware vSphere network type for the specified VMware cluster.
1.9.7. Managed cluster creation fails with an error processing disk changes
1.9.7.1. Symptom: Adding the VMware vSphere managed cluster fails due to an error processing disk changes
After creating a new Red Hat OpenShift Container Platform cluster on VMware vSphere, the cluster fails because there is an error when processing disk changes.
1.9.7.2. Identifying the problem: Adding the VMware vSphere managed cluster fails due to an error processing disk changes
A message similar to the following is displayed in the logs:
ERROR ERROR Error: error reconfiguring virtual machine: error processing disk changes post-clone: disk.0: ServerFaultCode: NoPermission: RESOURCE (vm-71:2000), ACTION (queryAssociatedProfile): RESOURCE (vm-71), ACTION (PolicyIDByVirtualDisk)
1.9.7.3. Resolving the problem: Adding the VMware vSphere managed cluster fails due to an error processing disk changes
Use the VMware vSphere client to give the user All privileges for Profile-driven Storage Privileges.
1.10. Managed cluster creation fails on Red Hat OpenStack Platform with unknown authority error
If you experience a problem when creating a Red Hat OpenShift Container Platform cluster on Red Hat OpenStack Platform, see the following troubleshooting information to see if one of them addresses your problem.
1.10.1. Symptom: Managed cluster creation fails with unknown authority error
After creating a new Red Hat OpenShift Container Platform cluster on Red Hat OpenStack Platform using self-signed certificates, the cluster fails with an error message that indicates an unknown authority error.
1.10.2. Identifying the problem: Managed cluster creation fails with unknown authority error
The deployment of the managed cluster fails and returns the following error message:
x509: certificate signed by unknown authority
1.10.3. Resolving the problem: Managed cluster creation fails with unknown authority error
Verify that the following files are configured correctly:
The
clouds.yaml
file must specify the path to theca.crt
file in thecacert
parameter. Thecacert
parameter is passed to the OpenShift installer when generating the ignition shim. See the following example:clouds: openstack: cacert: "/etc/pki/ca-trust/source/anchors/ca.crt"
The
certificatesSecretRef
paremeter must reference a secret with a file name matching theca.crt
file. See the following example:spec: baseDomain: dev09.red-chesterfield.com clusterName: txue-osspoke platform: openstack: cloud: openstack credentialsSecretRef: name: txue-osspoke-openstack-creds certificatesSecretRef: name: txue-osspoke-openstack-certificatebundle
To create a secret with a matching file name, run the following command:
oc create secret generic txue-osspoke-openstack-certificatebundle --from-file=ca.crt=ca.crt.pem -n $CLUSTERNAME
-
The size of the
ca.cert
file must be less than 63 thousand bytes.
1.11. Troubleshooting OpenShift Container Platform version 3.11 cluster import failure
1.11.1. Symptom: OpenShift Container Platform version 3.11 cluster import failure
After you attempt to import a Red Hat OpenShift Container Platform version 3.11 cluster, the import fails with a log message that resembles the following content:
customresourcedefinition.apiextensions.k8s.io/klusterlets.operator.open-cluster-management.io configured clusterrole.rbac.authorization.k8s.io/klusterlet configured clusterrole.rbac.authorization.k8s.io/open-cluster-management:klusterlet-admin-aggregate-clusterrole configured clusterrolebinding.rbac.authorization.k8s.io/klusterlet configured namespace/open-cluster-management-agent configured secret/open-cluster-management-image-pull-credentials unchanged serviceaccount/klusterlet configured deployment.apps/klusterlet unchanged klusterlet.operator.open-cluster-management.io/klusterlet configured Error from server (BadRequest): error when creating "STDIN": Secret in version "v1" cannot be handled as a Secret: v1.Secret.ObjectMeta: v1.ObjectMeta.TypeMeta: Kind: Data: decode base64: illegal base64 data at input byte 1313, error found in #10 byte of ...|dhruy45="},"kind":"|..., bigger context ...|tye56u56u568yuo7i67i67i67o556574i"},"kind":"Secret","metadata":{"annotations":{"kube|...
1.11.2. Identifying the problem: OpenShift Container Platform version 3.11 cluster import failure
This often occurs because the installed version of the kubectl
command-line tool is 1.11, or earlier. Run the following command to see which version of the kubectl
command-line tool you are running:
kubectl version
If the returned data lists version 1.11, or earlier, complete one of the fixes in Resolving the problem: OpenShift Container Platform version 3.11 cluster import failure.
1.11.3. Resolving the problem: OpenShift Container Platform version 3.11 cluster import failure
You can resolve this issue by completing one of the following procedures:
Install the latest version of the
kubectl
command-line tool.-
Download the latest version of the
kubectl
tool from: Install and Set Up kubectl in the Kubernetes documentation. -
Import the cluster again after upgrading your
kubectl
tool.
-
Download the latest version of the
Run a file that contains the import command.
- Start the procedure in Importing a managed cluster with the CLI.
-
When you create the command to import your cluster, copy that command into a YAML file named
import.yaml
. Run the following command to import the cluster again from the file:
oc apply -f import.yaml
1.12. Troubleshooting imported clusters offline after certificate change
Installing a custom apiserver
certificate is supported, but one or more clusters that were imported before you changed the certificate information can have an offline
status.
1.12.1. Symptom: Clusters offline after certificate change
After you complete the procedure for updating a certificate secret, one or more of your clusters that were online are now displaying an offline
status in the Red Hat Advanced Cluster Management for Kubernetes console.
1.12.2. Identifying the problem: Clusters offline after certificate change
After updating the information for a custom API server certificate, clusters that were imported and running before the new certificate are now in an offline
state.
The errors that indicate that the certificate is the problem are found in the logs for the pods in the open-cluster-management-agent
namespace of the offline managed cluster. The following examples are similar to the errors that are displayed in the logs:
Log of work-agent
:
E0917 03:04:05.874759 1 manifestwork_controller.go:179] Reconcile work test-1-klusterlet-addon-workmgr fails with err: Failed to update work status with err Get "https://api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/namespaces/test-1/manifestworks/test-1-klusterlet-addon-workmgr": x509: certificate signed by unknown authority E0917 03:04:05.874887 1 base_controller.go:231] "ManifestWorkAgent" controller failed to sync "test-1-klusterlet-addon-workmgr", err: Failed to update work status with err Get "api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/namespaces/test-1/manifestworks/test-1-klusterlet-addon-workmgr": x509: certificate signed by unknown authority E0917 03:04:37.245859 1 reflector.go:127] k8s.io/client-go@v0.19.0/tools/cache/reflector.go:156: Failed to watch *v1.ManifestWork: failed to list *v1.ManifestWork: Get "api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/namespaces/test-1/manifestworks?resourceVersion=607424": x509: certificate signed by unknown authority
Log of registration-agent
:
I0917 02:27:41.525026 1 event.go:282] Event(v1.ObjectReference{Kind:"Namespace", Namespace:"open-cluster-management-agent", Name:"open-cluster-management-agent", UID:"", APIVersion:"v1", ResourceVersion:"", FieldPath:""}): type: 'Normal' reason: 'ManagedClusterAvailableConditionUpdated' update managed cluster "test-1" available condition to "True", due to "Managed cluster is available" E0917 02:58:26.315984 1 reflector.go:127] k8s.io/client-go@v0.19.0/tools/cache/reflector.go:156: Failed to watch *v1beta1.CertificateSigningRequest: Get "https://api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/managedclusters?allowWatchBookmarks=true&fieldSelector=metadata.name%3Dtest-1&resourceVersion=607408&timeout=9m33s&timeoutSeconds=573&watch=true"": x509: certificate signed by unknown authority E0917 02:58:26.598343 1 reflector.go:127] k8s.io/client-go@v0.19.0/tools/cache/reflector.go:156: Failed to watch *v1.ManagedCluster: Get "https://api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/managedclusters?allowWatchBookmarks=true&fieldSelector=metadata.name%3Dtest-1&resourceVersion=607408&timeout=9m33s&timeoutSeconds=573&watch=true": x509: certificate signed by unknown authority E0917 02:58:27.613963 1 reflector.go:127] k8s.io/client-go@v0.19.0/tools/cache/reflector.go:156: Failed to watch *v1.ManagedCluster: failed to list *v1.ManagedCluster: Get "https://api.aaa-ocp.dev02.location.com:6443/apis/cluster.management.io/v1/managedclusters?allowWatchBookmarks=true&fieldSelector=metadata.name%3Dtest-1&resourceVersion=607408&timeout=9m33s&timeoutSeconds=573&watch=true"": x509: certificate signed by unknown authority
1.12.3. Resolving the problem: Clusters offline after certificate change
If your managed cluster is the local-cluster
or your managed cluster was created by using Red Hat Advanced Cluster Management for Kubernetes, you must wait 10 minutes or longer to reimport your managed cluster.
To reimport your managed cluster immediately, you can delete your managed cluster import secret on the hub cluster and reimport it by using Red Hat Advanced Cluster Management. Run the following command:
oc delete secret -n <cluster_name> <cluster_name>-import
Replace <cluster_name>
with the name of the managed cluster that you want to import.
If you want to reimport a managed cluster that was imported by using Red Hat Advanced Cluster Management, complete the following steps to import the managed cluster again:
On the hub cluster, recreate the managed cluster import secret by running the following command:
oc delete secret -n <cluster_name> <cluster_name>-import
Replace
<cluster_name>
with the name of the managed cluster that you want to import.On the hub cluster, expose the managed cluster import secret to a YAML file by running the following command:
oc get secret -n <cluster_name> <cluster_name>-import -ojsonpath='{.data.import\.yaml}' | base64 --decode > import.yaml
Replace
<cluster_name>
with the name of the managed cluster that you want to import.On the managed cluster, apply the
import.yaml
file by running the following command:oc apply -f import.yaml
Note: The previous steps do not detach the managed cluster from the hub cluster. The steps update the required manifests with current settings on the managed cluster, including the new certificate information.
1.13. Namespace remains after deleting a cluster
When you remove a managed cluster, the namespace is normally removed as part of the cluster removal process. In rare cases, the namespace remains with some artifacts in it. In that case, you must manually remove the namespace.
1.13.1. Symptom: Namespace remains after deleting a cluster
After removing a managed cluster, the namespace is not removed.
1.13.2. Resolving the problem: Namespace remains after deleting a cluster
Complete the following steps to remove the namespace manually:
Run the following command to produce a list of the resources that remain in the <cluster_name> namespace:
oc api-resources --verbs=list --namespaced -o name | grep -E '^secrets|^serviceaccounts|^managedclusteraddons|^roles|^rolebindings|^manifestworks|^leases|^managedclusterinfo|^appliedmanifestworks'|^clusteroauths' | xargs -n 1 oc get --show-kind --ignore-not-found -n <cluster_name>
Replace
cluster_name
with the name of the namespace for the cluster that you attempted to remove.Delete each identified resource on the list that does not have a status of
Delete
by entering the following command to edit the list:oc edit <resource_kind> <resource_name> -n <namespace>
Replace
resource_kind
with the kind of the resource. Replaceresource_name
with the name of the resource. Replacenamespace
with the name of the namespace of the resource.-
Locate the
finalizer
attribute in the in the metadata. -
Delete the non-Kubernetes finalizers by using the vi editor
dd
command. -
Save the list and exit the
vi
editor by entering the:wq
command. Delete the namespace by entering the following command:
oc delete ns <cluster-name>
Replace
cluster-name
with the name of the namespace that you are trying to delete.
1.14. Auto-import-secret-exists error when importing a cluster
Your cluster import fails with an error message that reads: auto import secret exists.
1.14.1. Symptom: Auto import secret exists error when importing a cluster
When importing a hive cluster for management, an auto-import-secret already exists
error is displayed.
1.14.2. Resolving the problem: Auto-import-secret-exists error when importing a cluster
This problem occurs when you attempt to import a cluster that was previously managed by Red Hat Advanced Cluster Management. When this happens, the secrets conflict when you try to reimport the cluster.
To work around this problem, complete the following steps:
To manually delete the existing
auto-import-secret
, run the following command on the hub cluster:oc delete secret auto-import-secret -n <cluster-namespace>
Replace
cluster-namespace
with the namespace of your cluster.- Import your cluster again using the procedure in Importing a target managed cluster to a hub cluster.
Your cluster is imported.
1.15. Troubleshooting cluster status changing from offline to available
The status of the managed cluster alternates between offline
and available
without any manual change to the environment or cluster.
1.15.1. Symptom: Cluster status changing from offline to available
When the network that connects the managed cluster to the hub cluster is unstable, the status of the managed cluster that is reported by the hub cluster cycles between offline
and available
.
The connection between the hub cluster and managed cluster is maintained through a lease that is validated at the leaseDurationSeconds
interval value. If the lease is not validated within five consecutive attempts of the leaseDurationSeconds
value, then the cluster is marked offline
.
For example, the cluster is marked offline
after five minutes with a leaseDurationSeconds
interval of 60 seconds
. This configuration can be inadequate for reasons such as connectivity issues or latency, causing instability.
1.15.2. Resolving the problem: Cluster status changing from offline to available
The five validation attempts is default and cannot be changed, but you can change the leaseDurationSeconds
interval.
Determine the amount of time, in minutes, that you want the cluster to be marked as offline
, then multiply that value by 60 to convert to seconds. Then divide by the default five number of attempts. The result is your leaseDurationSeconds
value.
Edit your
ManagedCluster
specification on the hub cluster by entering the following command, but replacecluster-name
with the name of your managed cluster:oc edit managedcluster <cluster-name>
Increase the value of
leaseDurationSeconds
in yourManagedCluster
specification, as seen in the following sample YAML:apiVersion: cluster.open-cluster-management.io/v1 kind: ManagedCluster metadata: name: <cluster-name> spec: hubAcceptsClient: true leaseDurationSeconds: 60
- Save and apply the file.
1.16. Troubleshooting cluster in console with pending or failed status
If you observe Pending status or Failed status in the console for a cluster you created, follow the procedure to troubleshoot the problem.
1.16.1. Symptom: Cluster in console with pending or failed status
After creating a new cluster by using the Red Hat Advanced Cluster Management for Kubernetes console, the cluster does not progress beyond the status of Pending or displays Failed status.
1.16.2. Identifying the problem: Cluster in console with pending or failed status
If the cluster displays Failed status, navigate to the details page for the cluster and follow the link to the logs provided. If no logs are found or the cluster displays Pending status, continue with the following procedure to check for logs:
Procedure 1
Run the following command on the hub cluster to view the names of the Kubernetes pods that were created in the namespace for the new cluster:
oc get pod -n <new_cluster_name>
Replace
new_cluster_name
with the name of the cluster that you created.If no pod that contains the string
provision
in the name is listed, continue with Procedure 2. If there is a pod withprovision
in the title, run the following command on the hub cluster to view the logs of that pod:oc logs <new_cluster_name_provision_pod_name> -n <new_cluster_name> -c hive
Replace
new_cluster_name_provision_pod_name
with the name of the cluster that you created, followed by the pod name that containsprovision
.- Search for errors in the logs that might explain the cause of the problem.
Procedure 2
If there is not a pod with
provision
in its name, the problem occurred earlier in the process. Complete the following procedure to view the logs:Run the following command on the hub cluster:
oc describe clusterdeployments -n <new_cluster_name>
Replace
new_cluster_name
with the name of the cluster that you created. For more information about cluster installation logs, see Gathering installation logs in the Red Hat OpenShift documentation.- See if there is additional information about the problem in the Status.Conditions.Message and Status.Conditions.Reason entries of the resource.
1.16.3. Resolving the problem: Cluster in console with pending or failed status
After you identify the errors in the logs, determine how to resolve the errors before you destroy the cluster and create it again.
The following example provides a possible log error of selecting an unsupported zone, and the actions that are required to resolve it:
No subnets provided for zones
When you created your cluster, you selected one or more zones within a region that are not supported. Complete one of the following actions when you recreate your cluster to resolve the issue:
- Select a different zone within the region.
- Omit the zone that does not provide the support, if you have other zones listed.
- Select a different region for your cluster.
After determining the issues from the log, destroy the cluster and recreate it.
See Creating a cluster for more information about creating a cluster.
1.17. Troubleshooting application Git server connection
Logs from the open-cluster-management
namespace display failure to clone the Git repository.
1.17.1. Symptom: Git server connection
The logs from the subscription controller pod multicluster-operators-hub-subscription-<random-characters>
in the open-cluster-management
namespace indicates that it fails to clone the Git repository. You receive a x509: certificate signed by unknown authority
error, or BadGateway
error.
1.17.2. Resolving the problem: Git server connection
Important: Upgrade if you are on a previous version.
- Save apps.open-cluster-management.io_channels_crd.yaml as the same file name.
On the Red Hat Advanced Cluster Management cluster, run the following command to apply the file:
oc apply -f apps.open-cluster-management.io_channels_crd.yaml
In the
open-cluster-management
namespace, edit theadvanced-cluster-management.<version, example 2.5.0>
CSV, run the following command and edit:oc edit csv advanced-cluster-management.<version, example 2.5.0> -n open-cluster-management
Find the following containers:
-
multicluster-operators-standalone-subscription
multicluster-operators-hub-subscription
Replace the container images with the container that you want to use:
quay.io/open-cluster-management/multicluster-operators-subscription:<your image tag>
The update recreates the following pods in the
open-cluster-management
namespace:-
multicluster-operators-standalone-subscription-<random-characters>
-
multicluster-operators-hub-subscription-<random-characters>
-
Check that the new pods are running with the new docker image. Run the following command, then find the new docker image:
oc get pod multicluster-operators-standalone-subscription-<random-characters> -n open-cluster-management -o yaml oc get pod multicluster-operators-hub-subscription-<random-characters> -n open-cluster-management -o yaml
Update the images on managed clusters.
On the hub cluster, run the following command to update the image value in the
multicluster_operators_subscription
key to the image that you want to use:oc edit configmap -n open-cluster-management mch-image-manifest-<version, example 2.5.0> ... data: multicluster_operators_subscription: <your image with tag>
Restart the existing
multicluster-operators-hub-subscription
pod:oc delete pods -n open-cluster-management multicluster-operators-hub-subscription--<random-characters>
This recreates the
application-manager-<random-characters>
pod inopen-cluster-management-agent-addon
namespace on the managed cluster.- Check that the new pod is running with the new docker image.
When you create an application through the console or the CLI, add `insecureSkipVerify: true' in the channel spec manually. See the following example:
apiVersion: apps.open-cluster-management.io/v1 kind: Channel metadata: labels: name: sample-channel namespace: sample spec: type: GitHub pathname: <Git URL> insecureSkipVerify: true
1.18. Troubleshooting Grafana
When you query some time-consuming metrics in the Grafana explorer, you might encounter a Gateway Time-out
error.
1.18.1. Symptom: Grafana explorer gateway timeout
If you hit the Gateway Time-out
error when you query some time-consuming metrics in the Grafana explorer, it is possible that the timeout is caused by the Grafana in the open-cluster-management-observability
namespace.
1.18.2. Resolving the problem: Configure the Grafans
If you have this problem, complete the following steps:
Verify that the default configuration of Grafana has expected timeout settings:
To verify that the default timeout setting of Grafana, run the following command:
oc get secret grafana-config -n open-cluster-management-observability -o jsonpath="{.data.grafana\.ini}" | base64 -d | grep dataproxy -A 4
The following timeout settings should be displayed:
[dataproxy] timeout = 300 dial_timeout = 30 keep_alive_seconds = 300
To verify the default data source query timeout for Grafana, run the following command:
oc get secret/grafana-datasources -n open-cluster-management-observability -o jsonpath="{.data.datasources\.yaml}" | base64 -d | grep queryTimeout
The following timeout settings should be displayed:
queryTimeout: 300s
If you verified the default configuration of Grafana has expected timeout settings, then you can configure the Grafana in the
open-cluster-management-observability
namespace by running the following command:oc annotate route grafana -n open-cluster-management-observability --overwrite haproxy.router.openshift.io/timeout=300s
Refresh the Grafana page and try to query the metrics again. The Gateway Time-out
error is no longer displayed.
1.19. Troubleshooting local cluster not selected with placement rule
The managed clusters are selected with a placement rule, but the local-cluster
, which is a hub cluster that is also managed, is not selected. The placement rule user is not granted permission to get the managedcluster
resources in the local-cluster
namespace.
1.19.1. Symptom: Troubleshooting local cluster not selected as a managed cluster
All managed clusters are selected with a placement rule, but the local-cluster
is not. The placement rule user is not granted permission to get the managedcluster
resources in the local-cluster
namespace.
1.19.2. Resolving the problem: Troubleshooting local cluster not selected as a managed cluster
To resolve this issue, you need to grant the managedcluster
administrative permission in the local-cluster
namespace. Complete the following steps:
Confirm that the list of managed clusters does include
local-cluster
, and that the placement ruledecisions
list does not display thelocal-cluster
. Run the following command and view the results:% oc get managedclusters
See in the sample output that
local-cluster
is joined, but it is not in the YAML forPlacementRule
:NAME HUB ACCEPTED MANAGED CLUSTER URLS JOINED AVAILABLE AGE local-cluster true True True 56d cluster1 true True True 16h
apiVersion: apps.open-cluster-management.io/v1 kind: PlacementRule metadata: name: all-ready-clusters namespace: default spec: clusterSelector: {} status: decisions: - clusterName: cluster1 clusterNamespace: cluster1
Create a
Role
in your YAML file to grant themanagedcluster
administrative permission in thelocal-cluster
namespace. See the following example:apiVersion: rbac.authorization.k8s.io/v1 kind: Role metadata: name: managedcluster-admin-user-zisis namespace: local-cluster rules: - apiGroups: - cluster.open-cluster-management.io resources: - managedclusters verbs: - get
Create a
RoleBinding
resource to grant the placement rule user access to thelocal-cluster
namespace. See the following example:apiVersion: rbac.authorization.k8s.io/v1 kind: RoleBinding metadata: name: managedcluster-admin-user-zisis namespace: local-cluster roleRef: apiGroup: rbac.authorization.k8s.io kind: Role name: managedcluster-admin-user-zisis namespace: local-cluster subjects: - kind: User name: zisis apiGroup: rbac.authorization.k8s.io
1.20. Troubleshooting application Kubernetes deployment version
A managed cluster with a deprecated Kubernetes apiVersion
might not be supported. See the Kubernetes issue for more details about the deprecated API version.
1.20.1. Symptom: Application deployment version
If one or more of your application resources in the Subscription YAML file uses the deprecated API, you might receive an error similar to the following error:
failed to install release: unable to build kubernetes objects from release manifest: unable to recognize "": no matches for kind "Deployment" in version "extensions/v1beta1"
Or with new Kubernetes API version in your YAML file named old.yaml
for instance, you might receive the following error:
error: unable to recognize "old.yaml": no matches for kind "Deployment" in version "deployment/v1beta1"
1.20.2. Resolving the problem: Application deployment version
Update the
apiVersion
in the resource. For example, if the error displays for Deployment kind in the subscription YAML file, you need to update theapiVersion
fromextensions/v1beta1
toapps/v1
.See the following example:
apiVersion: apps/v1 kind: Deployment
Verify the available versions by running the following command on the managed cluster:
kubectl explain <resource>
-
Check for
VERSION
.
1.21. Troubleshooting Klusterlet with degraded conditions
The Klusterlet degraded conditions can help to diagnose the status of Klusterlet agents on managed cluster. If a Klusterlet is in the degraded condition, the Klusterlet agents on managed cluster might have errors that need to be troubleshooted. See the following information for Klusterlet degraded conditions that are set to True
.
1.21.1. Symptom: Klusterlet is in the degraded condition
After deploying a Klusterlet on managed cluster, the KlusterletRegistrationDegraded
or KlusterletWorkDegraded
condition displays a status of True.
1.21.2. Identifying the problem: Klusterlet is in the degraded condition
Run the following command on the managed cluster to view the Klusterlet status:
kubectl get klusterlets klusterlet -oyaml
-
Check
KlusterletRegistrationDegraded
orKlusterletWorkDegraded
to see if the condition is set toTrue
. Proceed to Resolving the problem for any degraded conditions that are listed.
1.21.3. Resolving the problem: Klusterlet is in the degraded condition
See the following list of degraded statuses and how you can attempt to resolve those issues:
-
If the
KlusterletRegistrationDegraded
condition with a status of True and the condition reason is: BootStrapSecretMissing, you need create a bootstrap secret onopen-cluster-management-agent
namespace. -
If the
KlusterletRegistrationDegraded
condition displays True and the condition reason is a BootstrapSecretError, or BootstrapSecretUnauthorized, then the current bootstrap secret is invalid. Delete the current bootstrap secret and recreate a valid bootstrap secret onopen-cluster-management-agent
namespace. -
If the
KlusterletRegistrationDegraded
andKlusterletWorkDegraded
displays True and the condition reason is HubKubeConfigSecretMissing, delete the Klusterlet and recreate it. -
If the
KlusterletRegistrationDegraded
andKlusterletWorkDegraded
displays True and the condition reason is: ClusterNameMissing, KubeConfigMissing, HubConfigSecretError, or HubConfigSecretUnauthorized, delete the hub cluster kubeconfig secret fromopen-cluster-management-agent
namespace. The registration agent will bootstrap again to get a new hub cluster kubeconfig secret. -
If the
KlusterletRegistrationDegraded
displays True and the condition reason is GetRegistrationDeploymentFailed, or UnavailableRegistrationPod, you can check the condition message to get the problem details and attempt to resolve. -
If the
KlusterletWorkDegraded
displays True and the condition reason is GetWorkDeploymentFailed ,or UnavailableWorkPod, you can check the condition message to get the problem details and attempt to resolve.
1.22. Troubleshooting Klusterlet application manager on managed clusters
When you upgrade from Red Hat Advanced Cluster Management for Kubernetes, the klusterlet-addon-appmgr
pod on Red Hat OpenShift Container Platform managed clusters version 4.5 and 4.6 are OOMKilled
.
1.22.1. Symptom: Klusterlet application manager on managed cluster
You receive an error for the klusterlet-addon-appmgr
pod on Red Hat OpenShift Container Platform managed clusters version 4.5 and 4.6: OOMKilled
.
1.22.2. Resolving the problem: Klusterlet application manager on managed cluster
For Red Hat Advanced Cluster Management for Kubernetes 2.1.x and 2.2, you need to manually increase the memory limit of the pod to 8Gb
. See the following steps.
On your hub cluster, annotate the
klusterletaddonconfig
to pause replication. See the following command:oc annotate klusterletaddonconfig -n ${CLUSTER_NAME} ${CLUSTER_NAME} klusterletaddonconfig-pause=true -- overwrite=true
On your hub cluster, scale down the
klusterlet-addon-operator
. See the following command:oc edit manifestwork ${CLUSTER_NAME}-klusterlet-addon-operator -n ${CLUSTER_NAME}
Find the
klusterlet-addon-operator
Deployment and addreplicas: 0
to the spec to scale down.- apiVersion: apps/v1 kind: Deployment metadata: labels: app: cluster1 name: klusterlet-addon-operator namespace: open-cluster-management-agent-addon spec: replicas: 0
On the managed cluster, the
open-cluster-management-agent-addon/klusterlet-addon-operator
pod will be terminated.Log in to the managed cluster to manually increase the memory limit in the
appmgr
pod.Run the following command:
% oc edit deployments -n open-cluster-management-agent-addon klusterlet-addon-appmgr
For example, if the limit is 5G, increase the limit to 8G.
resources: limits: memory: 2Gi -> 8Gi requests: memory: 128Mi -> 256Mi
1.23. Troubleshooting Object storage channel secret
If you change the SecretAccessKey
, the subscription of an Object storage channel cannot pick up the updated secret automatically and you receive an error.
1.23.1. Symptom: Object storage channel secret
The subscription of an Object storage channel cannot pick up the updated secret automatically. This prevents the subscription operator from reconciliation and deploys resources from Object storage to the managed cluster.
1.23.2. Resolving the problem: Object storage channel secret
You need to manually input the credentials to create a secret, then refer to the secret within a channel.
Annotate the subscription CR in order to generate a reconcile single to subscription operator. See the following
data
specification:apiVersion: apps.open-cluster-management.io/v1 kind: Channel metadata: name: deva namespace: ch-obj labels: name: obj-sub spec: type: ObjectBucket pathname: http://ec2-100-26-232-156.compute-1.amazonaws.com:9000/deva sourceNamespaces: - default secretRef: name: dev --- apiVersion: v1 kind: Secret metadata: name: dev namespace: ch-obj labels: name: obj-sub data: AccessKeyID: YWRtaW4= SecretAccessKey: cGFzc3dvcmRhZG1pbg==
Run
oc annotate
to test:oc annotate appsub -n <subscription-namespace> <subscription-name> test=true
After you run the command, you can go to the Application console to verify that the resource is deployed to the managed cluster. Or you can log in to the managed cluster to see if the application resource is created at the given namespace.
1.24. Troubleshooting observability
After you install the observability component, the component might be stuck and an Installing
status is displayed.
1.24.1. Symptom: MultiClusterObservability resource status stuck
If the observability status is stuck in an Installing
status after you install and create the Observability custom resource definition (CRD), it is possible that there is no value defined for the spec:storageConfig:storageClass
parameter. Alternatively, the observability component automatically finds the default storageClass
, but if there is no value for the storage, the component remains stuck with the Installing
status.
1.24.2. Resolving the problem: MultiClusterObservability resource status stuck
If you have this problem, complete the following steps:
Verify that the observability components are installed:
To verify that the
multicluster-observability-operator
, run the following command:kubectl get pods -n open-cluster-management|grep observability
To verify that the appropriate CRDs are present, run the following command:
kubectl get crd|grep observ
The following CRDs must be displayed before you enable the component:
multiclusterobservabilities.observability.open-cluster-management.io observabilityaddons.observability.open-cluster-management.io observatoria.core.observatorium.io
- If you create your own storageClass for a Bare Metal cluster, see How to create an NFS provisioner in the cluster or out of the cluster.
-
To ensure that the observability component can find the default storageClass, update the
storageClass
parameter in themulticluster-observability-operator
CRD. Your parameter might resemble the following value:
storageclass.kubernetes.io/is-default-class: "true"
The observability component status is updated to a Ready status when the installation is complete. If the installation fails to complete, the Fail status is displayed.
1.25. Troubleshooting OpenShift monitoring service
Observability service in a managed cluster needs to scrape metrics from the OpenShift Container Platform monitoring stack. The metrics-collector
is not installed if the OpenShift Container Platform monitoring stack is not ready.
1.25.1. Symptom: OpenShift monitoring service is not ready
The endpoint-observability-operator-x
pod checks if the prometheus-k8s
service is available in the openshift-monitoring
namespace. If the service is not present in the openshift-monitoring
namespace, then the metrics-collector
is not deployed. You might receive the following error message: Failed to get prometheus resource
.
1.25.2. Resolving the problem: OpenShift monitoring service is not ready
If you have this problem, complete the following steps:
- Log in to your OpenShift Container Platform cluster.
-
Access the
openshift-monitoring
namespace to verify that theprometheus-k8s
service is available. -
Restart
endpoint-observability-operator-x
pod in theopen-cluster-management-addon-observability
namespace of the managed cluster.
1.26. Troubleshooting metrics-collector
When the observability-client-ca-certificate
secret is not refreshed in the managed cluster, you might receive an internal server error.
1.28. Troubleshooting Submariner not connecting after installation - general information
If you configure Submariner and it does not run correctly, there are some things that you can do to identify the problem and resolve it.
1.28.1. Symptom: Submariner not connecting after installation - general information
Your Submariner network is not communicating after installation.
1.28.2. Identifying the problem: Submariner not connecting after installation - general information
If the network connectivity is not established after deploying Submariner, begin the troubleshooting steps. Note that it might take several minutes for the processes to complete when you deploy Submariner.
1.28.3. Resolving the problem: Submariner not connecting after installation - general information
When Submariner does not run correctly after deployment, there are a few troubleshooting steps and resources that you can use to diagnose the problem:
Check for the following requirements to determine whether the components of Submariner deployed correctly:
-
The
submariner-addon
pod is running in theopen-cluster-management
namespace of your hub cluster. The following pods are running in the
submariner-operator
namespace of each managed cluster:- submariner-addon
- submariner-gateway
- submariner-routeagent
- submariner-operator
- submariner-globalnet (only if Globalnet is enabled in the ClusterSet)
- submariner-lighthouse-agent
- submariner-lighthouse-coredns
-
submariner-networkplugin-syncer (only if the specified CNI value is
OVNKubernetes
) - submariner-metrics-proxy
-
The
-
Run the
subctl diagnose
command to check the status of the required pods, with the exception of thesubmariner-addon
pods. -
Run the
subctl gather
command on the managed cluster to gather logs of various Submariner pods, with the exception of thesubmariner-addon
pods. Open an issue. If the other steps do not identify the problem, open an issue with the following information:
- Run subctl gather to collect all the relevant Submariner logs and add them to the issue.
-
Capture the information for the
submariner
instance of theManagedClusterAddon
resource type, and for thesubmariner
instance of theSubmarinerConfig
resource type from theManagedCluster
namespace on the hub cluster. Provide the following information in your issue, as well as the other template information:
- What happened?
- What you expected to happen?
- How do you reproduce it (as minimally and precisely as possible)?
- Anything else that we need to know?
Environment information:
-
Submariner version (use the
subctl version
command) -
Kubernetes version (use the
kubectl version
command) -
Diagnose information gathered (use the
subctl diagnose all
command) -
Gather information (use the
subctl gather
command) Cloud provider or hardware configuration:
-
OS (use the
cat /etc/os-release
command) -
Kernel (use the
uname -a
command)
-
OS (use the
- Install tools
- Other environment information that might be useful
-
Submariner version (use the
1.29. Troubleshooting Submariner add-on status is degraded
After adding the Submariner add-on to the clusters in your cluster set, the status in the Connection status, Agent status, and Gateway nodes show unexpected status for the clusters.
1.29.1. Symptom: Submariner add-on status is degraded
You add the Submariner add-on to the clusters in your cluster set, the following status is shown in the Gateway nodes, Agent status, and Connection status for the clusters:
Gateway nodes labeled
-
Progressing
: The process to label the gateway nodes started. -
Nodes not labeled
: The gateway nodes are not labeled, possibly because the process to label them has not completed. -
Nodes not labeled
: The gateway nodes are not yet labeled, possibly because the process is waiting for another process to finish. - Nodes labeled: The gateway nodes have been labeled.
-
Agent status
- Progressing: The installation of the Submariner agent started.
- Degraded: The Submariner agent is not running correctly, possibly because it is still in progress.
Connection status
- Progressing: The process to establish a connection with the Submariner add-on started.
-
Degraded: The connection is not ready. If you just installed the add-on, the process might still be in progress. If it was after the connection has already been established and running, then two clusters have lost the connection to each other. When there are multiple clusters, all clusters display a
Degraded
status if any of the clusters is in adisconnected state.
It will also show which clusters are connected, and which ones are disconnected.
1.29.2. Resolving the problem: Submariner add-on status is degraded
- The degraded status often resolves itself as the processes complete. You can see the current step of the process by clicking the status in the table. You can use that information to determine whether the process is finished and you need to take other troubleshooting steps.
For an issue that does not resolve itself, complete the following steps to troubleshoot the problem:
You can use the
diagnose
command with thesubctl
utility to run some tests on the Submariner connections when the following conditions exist:-
The Agent status or Connection status is in a
Degraded
state. Thediagnose
command provides detailed analysis about the issue. Everything is green in console, but the networking connections are not working correctly. The
diagnose
command helps to confirm that there are no other connectivity or deployment issues outside of the console. It is considered best practice to run thediagnostics
command after any deployment to identify issues.See
diagnose
in the Submariner for more information about how to run the command.
-
The Agent status or Connection status is in a
If a problem continues with the
Connection status
, you can start by running thediagnose
command of thesubctl
utility tool to get a more detailed status for the connection between two Submariner clusters. The format for the command is:subctl diagnose all --kubeconfig <path-to-kubeconfig-file>
Replace
path-to-kubeconfig-file
with the path to thekubeconfig
file. Seediagnose
in the Submariner documentation for more information about the command.Check the firewall settings. Sometimes a problem with the connection is caused by firewall permissions issues that prevent the clusters from communicating. This can cause the
Connection status
to show as degraded. Run the following command to check the firewall issues:subctl diagnose firewall inter-cluster <path-to-local-kubeconfig> <path-to-remote-cluster-kubeconfig>
Replace
path-to-local-kubeconfig
with the path to thekubeconfig
file of one of the clusters.Replace
path-to-remote-kubeconfig
with the path to thekubeconfig
file of the other cluster. you can run theverify
command with yoursubctl
utility tool to test the connection between two Submariner clusters. The basic format for the command is:If a problem continues with the
Connection status
, you can run theverify
command with yoursubctl
utility tool to test the connection between two Submariner clusters. The basic format for the command is:subctl verify --kubecontexts <cluster1>,<cluster2> [flags]
Replace
cluster1
andcluster2
with the names of the clusters that you are testing. Seeverify
in the Submariner documentation for more information about the command.After the troubleshooting steps resolve the issue, use the
benchmark
command with thesubctl
tool to establish a base on which to compare when you run additional diagnostics.See
benchmark
in the Submariner documentation for additional information about the options for the command.
1.30. Troubleshooting hub cluster metrics missing
Hub cluster metrics might not be collected.
1.30.1. Symptom: Metrics for hub cluster is missing
If the metrics service is missing on the hub cluster, then metrics are not collected.
1.30.2. Identifying the problem: Metric for hub cluster is missing
Check for the metrics service on your hub cluster with the following command:
oc get svc -n open-cluster-management hub-subscription-metrics
If the hub-subscription-metrics
is not listed, you must manually add the metric service.
1.30.3. Resolving the problem: Metric for hub cluster is missing
As a cluster administrator, create the metrics service for your hub cluster. Complete the following steps:
Create the metrics
Service
resource. Add your metricsService
resource for your hub cluster. Your YAML might resemble the following file:apiVersion: v1 kind: Service metadata: labels: app: hub-subscription-metrics name: hub-subscription-metrics namespace: open-cluster-management spec: ports: - name: metrics port: 8381 protocol: TCP targetPort: 8381 selector: app: multicluster-operators-hub-subscription sessionAffinity: None type: ClusterIP
Run the following command to apply the file:
oc apply -f
Verify that the metrics service is created. Run the following command:
oc get svc -n open-cluster-management hub-subscription-metrics
The metrics service should appear on your standalone cluster.
1.31. Troubleshooting managed cluster metrics missing
Managed cluster metrics might not be collected.
1.31.1. Symptom: Metric for managed cluster is missing
If the metrics service is missing on the managed cluster, then metrics are not collected.
1.31.2. Identifying the problem: Metric for managed cluster is missing
Check for the metrics service on the managed cluster with the following command:
oc get svc -n open-cluster-management mc-subscription-metrics
If the mc-subscription-metrics
is not listed, you must manually add the metric service.
1.31.3. Resolving the problem: Metric for managed cluster is missing
As a subscription operator, manually create the metrics service for your managed cluster. Complete the following steps:
Create and apply the metrics
Service
resource for the managed cluster. Add your metricsService
resource for your managed cluster. Your YAML might resemble the following file:apiVersion: v1 kind: Service metadata: labels: app: mc-subscription-metrics name: mc-subscription-metrics namespace: open-cluster-management-agent-addon spec: ports: - name: metrics port: 8388 protocol: TCP targetPort: 8388 selector: component: application-manager sessionAffinity: None type: ClusterIP
Run the following command to start a new YAML file:
oc apply -f
Verify that the metrics service is created. Run the following command:
oc get svc -n open-cluster-management mc-subscription-metrics
The metrics service should appear on your standalone cluster.
1.32. Troubleshooting standalone cluster metrics missing
Standalone cluster metrics might not be collected.
1.32.1. Symptom: Metric for standalone cluster is missing
If the metrics service is missing on your standalone cluster, then metrics are not collected.
1.32.2. Identifying the problem: Metric for standalone cluster is missing
Check for the metrics service on your standalone cluster. Run the following command:
oc get svc -n open-cluster-management standalone-subscription-metrics
If the standalone-subscription-metrics
is not listed, you must manually add the metric service.
1.32.3. Resolving the problem: Metric for standalone cluster is missing
As a subscription operator, manually create the metrics service for your standalone cluster. Complete the following steps:
Create the metrics
Service
resource for your standalone cluster. Add your metricsService
resource for your managed cluster. Your YAML might resemble the following file:apiVersion: v1 kind: Service metadata: labels: app: standalone-subscription-metrics name: standalone-subscription-metrics namespace: open-cluster-management spec: ports: - name: metrics port: 8389 protocol: TCP targetPort: 8389 selector: app: multicluster-operators-standalone-subscription sessionAffinity: None type: ClusterIP
Run the following command to apply your file:
oc apply -f
After you add your new metrics service for your standalone cluster, verify that the metrics service is created. Run the following command:
oc get svc -n open-cluster-management standalone-subscription-metrics
The metrics service should appear on your standalone cluster.
1.33. Troubleshooting restore status shows FinishedWithErrors
After you restore a backup, resources are restored correctly but the Red Hat Advanced Cluster Management restore resource shows a FinishedWithErrors
status.
1.33.1. Symptom: restore status shows FinishedWithErrors
Red Hat Advanced Cluster Management shows a FinishedWithErrors
status and one or more of the Velero restore resources created by the Red Hat Advanced Cluster Management restore show a PartiallyFailed
status.
1.33.2. Resolving the problem: restore status shows FinishedWithErrors
If you restore from a backup that is empty, you can safely ignore the FinishedWithErrors
status.
Red Hat Advanced Cluster Management for Kubernetes restore shows a cumulative status for all Velero restore resources. If one status is PartiallyFailed
and the others are Completed
, the cumulative status you see is PartiallyFailed
to notify you that there is at least one issue.
To resolve the issue, check the status for all individual Velero restore resources with a PartiallyFailed
status and view the logs for more details. You can get the log from the object storage directly, or download it from the OADP Operator by using the DownloadRequest
custom resource.
To create a DownloadRequest
from the console, complete the following steps:
- Navigate to Operators > Installed Operators > Create DownloadRequest.
-
Select
BackupLog
as your Kind and follow the console instructions to complete theDownloadRequest
creation.
1.34. Generic resources are removed when restoring a hub cluster backup
When you restore a hub cluster backup and use the cleanupBeforeRestore: CleanupRestored
paramater on the Restore.cluster.open-cluster-management.io
resource, the resources created by the acm-resources-generic-schedule
backup might be removed.
1.34.1. Symptom: Generic resources are removed when restoring a hub cluster backup
Resources backed up in the acm-resources-generic-schedule
backup do not appear on the restored hub cluster. If you check the backup operator log, you see a message that resembles the following:
_2023-06-08T13:42:48.066572033Z 2023-06-08T13:42:48.066Z INFO Deleting resource DRPlacementControl [c1-helloworld-placement-1-drpc.c1-helloworld] {"controller": "restore", "controllerGroup": "cluster.open-cluster-management.io", "controllerKind": "Restore", "restore": {"name":"restore-acm","namespace":"open-cluster-management-backup"}
1.34.2. Resolving the problem: Generic resources are removed when restoring a hub cluster backup
The resources are removed if the following conditions occur:
-
You have resources backed up by the
acm-resources-generic-schedule
backup that do not match the Secret or ConfigMap resource type with acluster.open-cluster-management.io/backup
label. -
You run a restore that uses a
Restore.cluster.open-cluster-management.io
resource and you set thecleanupBeforeRestore:
value toCleanupRestored
. -
The latest Red Hat Advanced Cluster Management backup set does not contain an
acm-resources-schedule
backup, so an older version of the backup is selected. As a result, theacm-resources-schedule
backup has a different timestamp than theacm-resources-generic-schedule
backup. During the post restore operation, when the CleanRestore option is processed, all generic resources are cleaned up because they do not have the same timestamp as theacm-resources-schedule backup
. To resolve the issue, run the restore operation again and set thecleanupBeforeRestore:
value toNone
.`
1.35. Troubleshooting OADP custom resource definitions
The MultiClusterHub
resource overwrites Velero and the OADP custom resource definitions with resources that are shipped and supported by the backup and restore component, even if the cluster-backup
flag is set to false
.
1.35.1. Symptom: OADP custom resource definitions
If you try to install the OADP operator manually, your installation fails if the version you are installing does not use the same custom resource definitions that are installed by the backup and restore component.
1.35.2. Resolving the problem: OADP custom resource definitions
To temporarily work around the issue, you can pause the MultiClusterHub
reconciliation in test environments. Run the following command before installing the OADP operator manually, which prevents the MultiClusterHub
from updating the custom resource definitions installed by your OADP operator version:
oc annotate mch multiclusterhub -n open-cluster-management mch-pause=true
1.36. Troubleshooting multiline YAML parsing
When you want to use the fromSecret
function to add contents of a Secret
resource into a Route
resource, the contents are displayed incorrectly.
1.36.1. Symptom: Troubleshooting multiline YAML parsing
When the managed cluster and hub cluster are the same cluster the certificate data is redacted, so the contents are not parsed as a template JSON string. You might receive the following error messages:
message: >- [spec.tls.caCertificate: Invalid value: "redacted ca certificate data": failed to parse CA certificate: data does not contain any valid RSA or ECDSA certificates, spec.tls.certificate: Invalid value: "redacted certificate data": data does not contain any valid RSA or ECDSA certificates, spec.tls.key: Invalid value: "": no key specified]
1.36.2. Resolving the problem: Troubleshooting multiline YAML parsing
Configure your certificate policy to retrieve the hub cluster and managed cluster fromSecret
values. Use the autoindent
function to update your certificate policy with the following content:
tls: certificate: | {{ print "{{hub fromSecret "open-cluster-management" "minio-cert" "tls.crt" hub}}" | base64dec | autoindent }}