Known issues for the CDP Private Cloud Data Services Management Console

Known Issues in Management Console 1.5.1

OPSX-4560: ECS server restart failed in an air gap environment as it requires "yum install" from repo

ECS service restart may fail in an air gap environment with a yum download error.

Workaround:

Open the /opt/cloudera/cm-agent/service/ecs/rke.sh file with a text editor and remove the following line:

yum -y install iscsi-initiator-utils nfs-utils

OPSAPS-68558: [7.9.5->7.11.3.2] CM upgrade failed with BeanCreationException: Error creating bean with name 'com.cloudera.server.cmf.TrialState'

After upgrading the Cloudera Manager package, the Cloudera Manager Server does not start. An error about "Active Commands" is shown in the Cloudera Manager Server log.

This may happen when the Private Cloud Data Services Control Plane is actively issuing requests to Cloudera Manager while an upgrade is being performed.

Workaround:

Before upgrading Cloudera Manager make sure there are no active commands. If there are any active commands, wait for them to complete before starting a Cloudera Manager upgrade.

If Cloudera Manager restart fails after upgrade due to an active getClientConfig command, check the Cloudera Manager server log for a "There are 1 active commands of type GetClientConfigFiles" error. This may block a Cloudera Manager restart after upgrade. Use the following steps to resolve this issue:

1. Login to Cloudera Manager database.
2. Search for any active GetClientConfigFiles command in the COMMANDS table.

   UPDATE COMMANDS SET active=0,success=false,state='CANCELLED' where command_id=<command_id>;  

3. Delete these entries, including foreign key dependencies, in the following tables:
   • PROCESSES
   • PROCESSES_DETAIL
   • COMMANDS_DETAIL

   cm=> DELETE FROM COMMANDS where command_id=1546340765;
   ERROR:  update or delete on table "commands" violates foreign key constraint "fk_process_command" on table "processes"
   DETAIL:  Key (command_id)=(1546340765) is still referenced from table "processes".
   cm=>
   cm=> DELETE FROM processes where command_id=1546340765;
   ERROR:  update or delete on table "processes" violates foreign key constraint "fk_processes_detail_process" on table "processes_detail"
   DETAIL:  Key (process_id)=(1546340766) is still referenced from table "processes_detail".
   cm=>
   cm=>                                        ^
   cm=> DELETE FROM processes_detail where process_id=1546340766;
   DELETE 1
   cm=> DELETE FROM processes where command_id=1546340765;
   DELETE 1
   cm=> DELETE FROM COMMANDS where command_id=1546340765;
   ERROR:  update or delete on table "commands" violates foreign key constraint "fk_commands_detail_command" on table "commands_detail"
   DETAIL:  Key (command_id)=(1546340765) is still referenced from table "commands_detail".
   cm=>
   cm=> DELETE FROM commands_detail where command_id=1546340765;
   DELETE 1
   cm=> DELETE FROM COMMANDS where command_id=1546340765;
   DELETE 1 

4. Restart the Cloudera Manager server.

External metadata databases are no longer supported on OCP

As of CDP Private Cloud Data Services 1.5.1, external Control Plane metadata databases are no longer supported. New installs require the use of an embedded Control Plane database. Upgrades from CDP Private Cloud Data Services 1.4.1 or 1.5.0 to 1.5.1 are supported, but there is currently no migration path from a previous external Control Plane database to the embedded Control Plane database. Upgrades from 1.4.0 or 1.5.0 with external Control Plane metadata databases also require additional steps, which are described in the CDP Private Cloud Data Services 1.5.1 upgrade topics.

DOCS-18031: Nodes are in "Not Ready" status during Rolling Restart of ECS

During a rolling restart of ECS, nodes are in a "Not Ready" state, and the dmesg command returns the following error message on the applicable nodes.

[Tue Aug  8 16:30:50 2023] nfs: server 10.46.157.145 not responding, timed out
[Tue Aug  8 16:31:16 2023] nfs: server 10.46.157.145 not responding, timed out
[Tue Aug  8 16:31:30 2023] nfs: server 10.46.157.145 not responding, timed out 

Also, the df command may hang on these hosts.

Workaround:

Run the following commands on each unavailable node:

1. Find the NFS mounts:

   mount | grep "nfs"

2. Force unmount:

   umount -f <mount points found in Step 1 separated with a space>

OPSAPS-67214: Single Node | Restart Stability | Rolling start is failing with "global timeout reached: 10m0s, error when evicting pods"

For the ECS service, rolling restart is not applicable to a single node cluster.

Workaround:

Instead of a rolling restart, you should stop and start the ECS Service.

CDPVC-1098: How to refresh the YuniKorn configuration

Sometimes it is possible for the scheduler state to go out of sync from the cluster state. This may result in pods in Pending and ApplicationRejected states, with pod events showing Placement Rule related errors. To recover from this, you may need to refresh the YuniKorn configuration.

Workaround:

Follow the steps in Refreshing the YuniKorn configuration.

OPSAPS-67340: L1 runs failing as service monitor is in bad health state

Service Monitor (SMON) ends up in a bad health state after restarting the Cloudeara Manager (CM) server, reporting problems with descriptor and metric schema age, when Kerberos and CM SPNEGO authentication are both enabled.

Workaround:

Use the following steps to restart SMON each time a CM server restart is required:

1. Stop SMON
2. Restart the CM server
3. Start SMON

If the health status is already bad, a simple restart of SMON is sufficient.

DOCS-15855: Networking API is deprecated after upgrade to CDP Private Cloud Data Services 1.5.1 (K8s 1.24)

When the control plane is upgraded from 1.4.1 to 1.5.1, the Kubernates version changes to 1.24. The Livy pods running in existing Virtual Clusters (VCs) use a deprecated networking API for creating ingress for Spark driver pods. Because the old networking API is deprecated and does not exist in Kubernates 1.24, any new job run will not work for the existing VCs.

CDPQE-24295: Update docker client on docker.lab.eng.hortonworks machine

When you attempt to execute the Docker command to fetch the Cloudera-provided images into your air-gapped environment, you may encounter an issue where Docker pulls an incorrect version of the HAProxy image, especially if you are using an outdated Docker client. This situation arises due to the Cloudera registry containing images with multiple platform versions. Unfortunately, older Docker clients may lack the capability to retrieve the appropriate architecture version, such as amd64.

Workaround:

Update the Docker client. It has been demonstrated that Docker 20.10.5 and later versions have been successful in resolving this problem.

OPSX-4326: OCP upgrade from 1.5.0 to 1.5.1 – Restore is unsuccessful after upgrade

After upgrading CDP Private Cloud Data Services on OCP from 1.5.0 to 1.5.1, Restore using a 1.5.0 backup could not be performed successfully.

Workaround:

Make a backup of the OpenShift routes before upgrading to 1.5.1. If you need to restore the control plane on a CDP Private Cloud Data Services1.5.1 OpenShift cluster using a 1.5.0 backup, contact Cloudera Customer Support.

OPSX-4266: ECS upgrade from 1.5.0 to 1.5.1 is failing in Cadence schema update job

When upgrading from ECS 1.5.0 to 1.5.1, the CONTROL_PLANE_CANARY fails with the following error:

Firing alerts for Control Plane: Job did not complete in time, Job failed to complete. 

And the cdp-release-cdp-cadence-schema-update job fails.

Workaround:

Use the following steps to manually execute the job:

1. Export the job manifest into a file:

   kubectl get job cdp-release-cdp-cadence-schema-update -n <cdp> -o yaml > job.yaml

2. Delete the cdp-release-cdp-cadence-schema-update job:

   kubectl delete job cdp-release-cdp-cadence-schema-update -n <cdp>

3. Remove runtime information from the manifest, such as:

   resourceVersion
   uid
   selector
     matchLabels
       controller-uid
   labels
     controller-uid
   status section

4. Create the job:

   kubectl apply -f job.yaml

If the job still fails, contact Cloudera Support.

OPSX-4076:

When you delete an environment after the backup event, the restore operation for the backup does not bring up the environment.

Create the environment manually.

OPSX-4295:

The logs for the backups created in CDP Private Cloud Data Services 1.5.0 version do not appear after you upgrade to version 1.5.1.

OPSX-4024: CM truststore import into unified truststore should handle duplicate CommonNames

If multiple CA certificates with the exact same value for the Common Name field are present in the Cloudera Manager truststore when a Private Cloud Data Services cluster is installed, only one of them may be imported into the Data Services truststore. This may cause certificate errors if an incorrect/old certificate is imported.

Workaround:

Remove old certificates from the Cloudera Manager truststore, and ensure certificates have unique Common Names.

OPSX-2713: PVC ECS Installation: Failed to perform First Run of services

If an issue is encountered during the Install Control Plane step of the Containerized Cluster First Run, installation will be re-attempted infinitely rather than the command failing.

Workaround:

Because the control plane is installed and uninstalled in a continuous cycle, it is often possible to address the cause of the failure while the command is still running, at which point the next attempted installation should succeed. If this is not successful, abort the First Run command, delete the Containerized Cluster, address the cause of the failure, and then retry from the beginning of the Add Cluster wizard. Any nodes that are reused must be cleaned before re-attempting installation.

OPSX-3666: mlx_crud_app DB connection fails with error "unable to create connection: x509: certificate relies on legacy Common Name field, use SANs instead"

After upgrade, the mlx-crud-app fails with the error "unable to create connection: x509: certificate relies on legacy Common Name field, use SANs instead"

Workaround:

If you are upgrading from CDP Private Cloud Data Services 1.4.1 or 1.5.0 to 1.5.1, and you were previously using an external database, you must regenerate the DB certificate with SAN before upgrading to CDP Private Cloud Data Services 1.5.1.

OPSAPS-66166: FreeIPA cmadminrole needs more privileges for PvC+ after upgrade

After upgrade, the Cloudera Manager admin role may be missing the Host Administrators privilege in an upgraded cluster.

Workaround:

The cluster administrator should run the following command to manually add this privilege to the role.

ipa role-add-privilege <cmadminrole> --privileges="Host Administrators"

COMOPS-2822: OCP error x509: certificate signed by unknown authority

The error x509: certificate signed by unknown authority usually means that the Docker daemon that is used by Kubernetes on the managed cluster does not trust the self-signed certificate.

Workaround:

Usually the fix is to copy the certificate to the path below on all of the worker nodes in the cluster:

/etc/docker/certs.d/<your_registry_host_name>:<your_registry_host_port>/ca.crt

OPSX-4225: Upgrade failed as cadence pods are crashlooping post upgrade

When doing a fresh install of CDP Private Cloud Data Services 1.5.1, external metadata databases are no longer supported. Instead, the CDP Private Cloud Data Services installer will create an embedded database pod by default, which runs inside the Kubernetes cluster to host the databases required for installation.

If you are upgrading from CDP Private Cloud Data Services 1.4.1 or 1.5.0 to 1.5.1, and you were previously using an external database, you must run the following psql commands to create the required databases. You should also ensure that the two new databases are owned by the common database users known by the control plane.

CREATE DATABASE db-cadence;
CREATE DATABASE db-cadence-visibility;

OPSAPS-67046: Docker Server role fails to come up and returns a connection error during ECS upgrade

When upgrading from 1.4.1 to 1.5.1, a Docker server role can sometimes fail to come up and return the following error:

grpc: addrConn.createTransport failed to connect to {unix:///var/run/docker/containerd/containerd.sock <nil> 0 <nil>}. 
Err :connection error: desc = "transport: error while dialing: dial unix:///var/run/docker/containerd/containerd.sock: timeout". 
Reconnecting... module=grpc 
failed to start containerd: timeout waiting for containerd to start

This error appears in the stderr file of the command, and can be caused by a mismatch in the pid of containerd.

Workaround:

1. Ensure that the problematic Docker server role has been stopped.
2. Log in to the failing Docker server host.
3. Run the following commands:

   cd /var/run/docker/containerd/
   rm containerd.pid

4. Restart the Docker server role.

Longhorn-4212 Somehow the Rebuilding field inside volume.meta is set to true causing the volume to get stuck in attaching/detaching loop

This is a condition that can occur in ECS Longhorn storage.

Since the volume has only 1 replica in this case, we can:

1. Scale down the workload. The Longhorn volume will be detached.

2. Wait for the Longhorn volume to be detached.

3. SSH into the node that has the replica.

4. cd into the replica folder (for example, /longhorn/replicas/pvc-126d40e2-7bff-4679-a310-e444e84df267-1a5dc941).

5. Change the"Rebuilding" field from true to false in the volume.meta file.

6. Scale up the workload to attach the volume.

OPSX-3073 [ECS] First run command failed at setup storage step with error "Timed out waiting for local path storage to come up"

Pod stuck in pending state on host for a long time. Error in Role log related to CNI plugin:

Events:

   Type     Reason                  Age                   From     Message
  ----     ------                  ----                  ----     -------
  Warning  FailedCreatePodSandBox  3m5s (x269 over 61m)  kubelet  (combined from similar events): 
Failed to create pod sandbox: rpc error: code = Unknown desc = failed to setup network for sandbox 
"70427e9b26fb014750dfe4441fdfae96cb4d73e3256ff5673217602d503e806f": 
failed to find plugin "calico" in path [/opt/cni/bin] 

Delete the cni directory on the host failing to launch pods:

ssh root@ecs-ha1-p-7.vpc.cloudera.com rm -rf /var/lib/cni

Restart the canal pod running on that host:

kubectl get pods -n kube-system -o wide | grep  ecs-ha1-p-7.vpc.cloudera.com
kube-proxy-ecs-ha1-p-7.vpc.cloudera.com                 1/1     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
rke2-canal-llkc9                                        2/2     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
rke2-ingress-nginx-controller-dqtz8                     1/1     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
kubectl delete pod rke2-canal-llkc9 -n kube-system

Known Issues in Management Console 1.5.0

Longhorn-4212 Somehow the Rebuilding field inside volume.meta is set to true causing the volume to get stuck in attaching/detaching loop

This is a condition that can occur in ECS Longhorn storage.

Since the volume has only 1 replica in this case, we can:

1. Scale down the workload. The Longhorn volume will be detached.

2. Wait for the Longhorn volume to be detached.

3. SSH into the node that has the replica.

4. cd into the replica folder (for example, /longhorn/replicas/pvc-126d40e2-7bff-4679-a310-e444e84df267-1a5dc941).

5. Change the"Rebuilding" field from true to false in the volume.meta file.

6. Scale up the workload to attach the volume.

COMPX-13185 Upgrade from 1.4.1 to 1.5.0 failed - error: timed out waiting for the condition on jobs/helm-install-longhorn

Before ECS upgrade, you must update a specific ECS server node toleration explicitly to ensure a cleaner upgrade process.

Delete the cni directory on the host failing to launch pods:

ssh root@ecs-ha1-p-7.vpc.cloudera.com rm -rf /var/lib/cni

Before ECS upgrade, run the following commands on the ECS Server hosts:

TOLERATION='{"spec": { "template": {"spec": { "tolerations": [{ "effect": "NoSchedule","key": "node-role.kubernetes.io/control-plane","operator": "Exists" }]}}}}'

kubectl patch deployment/yunikorn-admission-controller -n yunikorn -p "$TOLERATION"

kubectl patch deployment/yunikorn-scheduler -n yunikorn -p "$TOLERATION"

OPSX-3073 [ECS] First run command failed at setup storage step with error "Timed out waiting for local path storage to come up"

Pod stuck in pending state on host for a long time. Error in Role log related to CNI plugin:

Events:

   Type     Reason                  Age                   From     Message
  ----     ------                  ----                  ----     -------
  Warning  FailedCreatePodSandBox  3m5s (x269 over 61m)  kubelet  (combined from similar events): 
Failed to create pod sandbox: rpc error: code = Unknown desc = failed to setup network for sandbox 
"70427e9b26fb014750dfe4441fdfae96cb4d73e3256ff5673217602d503e806f": 
failed to find plugin "calico" in path [/opt/cni/bin] 

Delete the cni directory on the host failing to launch pods:

ssh root@ecs-ha1-p-7.vpc.cloudera.com rm -rf /var/lib/cni

Restart the canal pod running on that host:

kubectl get pods -n kube-system -o wide | grep  ecs-ha1-p-7.vpc.cloudera.com
kube-proxy-ecs-ha1-p-7.vpc.cloudera.com                 1/1     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
rke2-canal-llkc9                                        2/2     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
rke2-ingress-nginx-controller-dqtz8                     1/1     Running     0          11h   10.65.52.51    ecs-ha1-p-7.vpc.cloudera.com   <none>           <none>
kubectl delete pod rke2-canal-llkc9 -n kube-system

Known Issues in Management Console 1.4.1

INSIGHT-2469: COE Insight from case 922848: Not able to connect to bit bucket

After installing CML on an ECS cluster, users were not able to connect the internal bitbucket repo.

Workaround:

In this case the MTU of the ECS virtual network interfaces were larger than that of host external interface, which may cause the network requests from ECS containers to get truncated.

The Container Network Interface (CNI) is a framework for dynamically configuring networking resources. CNI integrates smoothly with Kubenetes to enable the use of an overlay or underlay network to automatically configure the network between pods. Cloudera ECS uses Calico as the CNI network provider.

The MTU of the pods’ virtual network interface can be seen by running the ifconfig command.

The default MTU of the virtual network interfaces is 1450.

The MTU setting of the virtual interfaces is stored as a configmap in the kube-system namespace. To modify the MTU, edit the rke2-canal-config configmap.

$ /var/lib/rancher/rke2/bin/kubectl --kubeconfig 
 /etc/rancher/rke2/rke2.yaml --namespace kube-system 
 edit cm rke2-canal-config

Find the veth_mtu parameter in the YAML content. Modify the default value of 1450 to the required MTU size.

Next, restart the rke2-canal pods from the kube-system namespace. There will be rke2-canal pods for each ECS node.

After the pods are restarted, all subsequent new pods will use the new MTU setting. However, existing pods that are already running will remain on the old MTU setting. Restart all of the pods to apply the new MTU setting.

OPSAPS-67046: Docker Server role fails to come up and returns a connection error

A Docker server role can sometimes fail to come up and return the following error:

grpc: addrConn.createTransport failed to connect to {unix:///var/run/docker/containerd/containerd.sock <nil> 0 <nil>}. 
Err :connection error: desc = "transport: error while dialing: dial unix:///var/run/docker/containerd/containerd.sock: timeout". 
Reconnecting... module=grpc 
failed to start containerd: timeout waiting for containerd to start

This error appears in the stderr file of the command, and can be caused by a mismatch in the pid of containerd.

Workaround:

1. Ensure that the problematic Docker server role has been stopped.
2. Log in to the failing Docker server host.
3. Run the following commands:

   cd /var/run/docker/containerd/
   rm containerd.pid

4. Restart the Docker server role.

OPSX-1405: Able to create multiple CDP PVC Environments with the same name

If two users try to create an environment with the same name at the same time, it might result in an unusable environment.

Delete the environment and try again with only one user trying to create the environment.

OPSX-1412: Creating a new environment through the CDP CLI reports intermittently that "Environment name is not unique" even though it is unique

When multiple users try to create the same environment at the same time or use automation to create an environment with retries, create environment may fail on collision with a previous request to create an environment.

Delete the existing environment, wait 5 minutes, and try again.

OPSX-3323: Custom Log Redaction | String is not getting redacted from all places in diagnostic bundle

Custom redaction rule for URLs does not work.

Cloudera Data Engineering service fails to start due to Ozone

If the Ozone service is missing, misconfigured, or having other issues when an Environment is registered in the Management Console, CDE fails to start.

Workaround:

1. Correct the issues with the Ozone service.
2. Ensure that Ozone is running as expected.
3. Re-create the environment.
4. Create a new Cloudera Data Engineering service.

Known Issues in Management Console 1.4.0

Cloudera Data Engineering service fails to start due to Ozone

If the Ozone service is missing, misconfigured, or having other issues when an Environment is registered in the Management Console, CDE fails to start.

Workaround:

1. Correct the issues with the Ozone service.
2. Ensure that Ozone is running as expected.
3. Re-create the environment.
4. Create a new Cloudera Data Engineering service.

OPSX-2062: Platform not shown on the Compute Cluster UI tab

On CDP Private Console UI in ECS, when listing the compute clusters, the Platform field is empty (null) instead of displaying RKE as the Platform.

None.

OPSX-2713: ECS Installation: Failed to perform First Run of services.

If an issue is encountered during the Install Control Plane step of Containerized Cluster First Run, installation will be re-attempted infinitely rather than the command failing.

Since the control plane is installed and uninstalled in a continuous cycle, it is often possible to address the cause of the failure while the command is still running, at which point the next attempted installation should succeed. If this is not successful, abort the First Run command, delete the Containerized Cluster, address the cause of the failure, and retry from the beginning of the Add Cluster wizard. Any nodes that are re-used must be cleaned before re-attempting installation.

OPSX-735: Kerberos service should handle CM downtime

The Cloudera Manager Server in the base cluster must be running in order to generate Kerberos principals for Private Cloud. If there is downtime, you may observe Kerberos-related errors.

Resolve downtime on Cloudera Manager. If you encountered Kerberos errors, you can retry the operation (such as retrying creation of the Virtual Warehouse).

OPSX-1405: Able to create multiple CDP PVC Environments with the same name

If two users try to create an environment with the same name at the same time, it might result in an unusable environment.

Delete the environment and try again with only one user trying to create the environment.

OPSX-1412: Creating a new environment through the CDP CLI intermittently reports that, "Environment name is not unique" even though it is unique

When multiple users try to create the same environment at the same time or use automation to create an environment with retries, create environment may fail on collision with a previous request to create an environment.

Delete the existing environment, wait 5 minutes, and try again.

OPSX-2484: FileAlreadyExistsException during timestamp filtering

The timestamp filtering may result in FileAlreadyExistsException when there is a file with same name already existing in the tmp directory.

OPSX-2772: For Account Administrator user, update roles functionality should be disabled

An Account Administrator user holds the biggest set of privileges, and is not allowed to modify via current UI, even user try to modify permissions system doesn't support changing for account administrator.

Known Issues for Management Console 1.3.x and lower

Recover fast in case of a Node failures with ECS HA

When a node is deleted from cloud or made unavailable, it is observed that the it takes more than two minutes until the pods were rescheduled on another node.

It takes some time for the nodes to recover. Failure detection and pod-transitioning are not instantaneous.

Cloudera Manager 7.6.1 is not compatible with CDP Private Cloud Data Servicesversion 1.3.4.

You must use Cloudera Manager version 7.5.5 with this release.

CDP Private Cloud Data Services ECS Installation: Failed to perform First Run of services.

If an issue is encountered during the Install Control Plane step of Containerized Cluster First Run, installation will be re-attempted infinitely rather than the command failing.

Workaround: Since the control plane is installed and uninstalled in a continuous cycle, it is often possible to address the cause of the failure while the command is still running, at which point the next attempted installation should succeed. If this is not successful, abort the First Run command, delete the Containerized Cluster, address the cause of the failure, and retry from the beginning of the Add Cluster wizard. Any nodes that are re-used must be cleaned before re-attempting installation.

Environment creation through the CDP CLI fails when the base cluster includes Ozone

Problem: Attempt to create an environment using the CDP command-line interface fails in a CDP Private Cloud Data Services deployment when the Private Cloud Base cluster is in a degraded state and includes Ozone service.

Workaround: Stopping the Ozone service temporarily in the Private Cloud Base cluster during environment creation prevents the control plane from using Ozone as a logging destination, and avoids this issue.

Filtering the diagnostic data by time range might result in a FileAlreadyExistsException

Problem:Filtering the collected diagnostic data might result in a FileAlreadyExistsException if the /tmp directory already contains a file by that name.

There is currently no workaround for this issue.

Full cluster name does not display in the Register Environment Wizard

None

Kerberos service does not always handle Cloudera Manager downtime

Problem: The Cloudera Manager Server in the base cluster must be running to generate Kerberos principals for CDP Private Cloud. If there is downtime, you might observe Kerberos-related errors.

Resolve downtime issues on Cloudera Manager. If you encounter Kerberos errors, you can retry the concerned operation such as creating Virtual Warehouses.

Management Console allows registration of two environments of the same name

Problem: If two users attempt to register environments of the same name at the same time, this might result in an unusable environment.

Delete the environment and ensure that only one user attempts to register a new environment.

Not all images are pushed during upgrade

A retry of a failed upgrade intermittently fails at the Copy Images to Docker Registry step due to images not being found locally.

The failed images can be loaded manually (with a docker load), and the upgrade resumed. To identify which images need to be loaded take a look at the stderr file. The downloaded images are present in the Docker Data Directory.

The Environments page on the Management Console UI for an environment in a deployment using ECS does not display the platform name

Problem: When you view the details of an environment using the Management Console UI in a CDP Private Cloud Data Services deployment using ECS, the Platform field appears blank.

Use the relevant CDP CLI command from the environments module to view the required details.

Updating user roles for the admin user does not update privileges

In the Management Console, changing roles on the User Management page does not change privileges of the admin user.

None

Upgrade applies values that cannot be patched

If the size of a persistent volume claim in a Containerized Cluster is manually modified, subsequent upgrades of the cluster will fail.

Incorrect warning about stale Kerberos client configurations

If Cloudera Manager is configured to manage krb5.conf, ECS clusters may display a warning that they have stale Kerberos client configurations. Clicking on the warning may show an "Access denied" error.

No action is needed. ECS clusters do not require Kerberos client configurations to be deployed on those hosts.

Vault becomes sealed

If a host in an ECS cluster fails or restarts, the Vault may have become sealed. (You may see a Health Test alert in Cloudera Manager for the ECS service stating Vault instance is sealed.)

Unseal the Vault. In the Cloudera Manager Admin Console, go to the ECS service and click Actions > Unseal .