Duplicate AI for workload clusters can be created by TKGm/TMC auto discover
Workaround 1: To auto manage and avoid any duplication of the newly created workload cluster by Supervisor cluster, you need to delete the manually created workload adapter instance. This will also clear the metric history on pods/container objects for the manually created adapter instance.
Workaround 2: Stop the collection of auto created K8s adapter instances. Stopping the adapter instance will not cause any memory or CPU usage in the collector but prevents duplication and monitoring (which happens after deleting the auto-created workload AI).
Some Prometheus exporter metrics are propagated to parent objects as the PromQL matches the same metric
Some Prometheus exporter metrics are propagated to parent objects as the PromQL matches the same metric. So, the the same set of metrics are grouped under parent Kubernetes objects. For example: The Container metrics are grouped under Pod/Node/Namespace.
Workaround: None
If you upgrade from VMware vRealize Operations Management Pack for Kubernetes 1.0 to 1.2.1, the collection state displays a status called Not Collecting for all the adapter instances
This occurs because of the addition of new settings and credential types in the 1.1 version of VMware vRealize Operations Management Pack for Kubernetes.
Note: If you upgrade from VMware vRealize Operations Management Pack for Kubernetes 1.1 to 1.2.1, you do not have to complete the steps listed below.
Workaround: All the adapter instances must be deleted and recreated. This will lead to creation of new objects. However, you can retain the old objects to keep historical data intact.
From the main menu of vRealize Operations Manager, click Administration, and then in the left pane click Solutions.
From the Solutions page, select VMware vRealize Operations Management Pack for Kubernetes.
Click the Configure icon. The Manage Solution dialog box appears.
Select an adapter instance.
Click the Delete icon.
When the Confirmation dialog box appears and if you want to retain historical data, deselect the option Remove related objects.
Recreate the adapter instance by following steps provided in User Guide.
Repeat the above steps for all adapter instances.
During configuration, VMware vRealize Operations Management Pack for Kubernetes verifies if the cAdvisor service is accessible on every node. An error message similar to the following may appear: Unable to establish a valid connection to the target system. cAdvisor service on following nodes is either not reachable OR of a lower version than v2.1
The error occurs if the cAdvisor service is inaccessible or if the API version is lesser than 2.1. You may sometimes receive this error if the cAdvisor service temporarily throws a gateway error at the time of verification.
Workaround:
Verify if the cAdvisor service is up and running on the affected nodes and responds to API calls.
Verify if the API version of the cAdvisor service is later than 2.1. If not, deploy the latest version of the cAdvisor service.
If you have completed the above two steps, you can ignore the error message and continue to save the adapter instance.
Under recommendations, the Defined by column is displayed as KubernetesAdapter3
Under recommendations, the Defined by column is displayed as KubernetesAdapter3.
Deleting the VMware PKS adapter instance does not remove the Kubernetes adapter instances created by the VMware PKS adapter instance
When you delete the VMware PKS adapter instance, the Kubernetes adapter instances created by the VMware PKS adapter instance will not be removed.
Workaround : Manually delete the adapter instances related to the VMware PKS adapter instance
The Environment Overview dashboard does not display the relationship between the vCenter Hosts/Virtual Machines and the Kubernetes nodes
If the vRealize Operations Manager accesses the K8s through proxy, the vCenter adapter instance does not provide a provision to specify proxy. So, the Environment Overview dashboard may not display the relationship between the vCenter Hosts/Virtual Machines and the Kubernetes nodes.
Workaround: None
Data collection fails for the K8s adapters that are auto-configured by the VMware PKS adapter
The auto-configuredK8s adapter instances that presents the untrusted SSL certificates will have the collection status as 'Failed'.
Workaround: Manually accept the untrusted certificate for the auto-configured K8s adapter instances for which data collection has failed.
Adding the VMware PKS adapter will configure the K8s instances but does not create the vCenter adapter instances
Adding the VMware PKS adapter will configure the K8s instances but does not create the vCenter adapter instances or associate it with the vCenters that the Kubernetes cluster nodes are deployed in.
Workaround: Manually configure the vCenter adapter instances and then add the details to the K8s adapters that are auto-configured by VMware PKS.
Expired certificates are also auto accepted when the 'Auto-accept Kubernetes Cluster SSL Certificate' option is enabled
When you enable the 'Auto-accept Kubernetes Cluster SSL Certificate' under Advanced Settings, the expired certificates are also auto accepted and there is no prompt.
Workaround: None
Container File System base usage (MB) should be referred as file system base limit
The Container File System base usage (MB) should be referred as file system base limit. This is applicable only for Prometheus monitored containers.
Workaround: None
DiskIO|Sync and DiskIO|Async Container metrics will not be shown in vRealize Operations Manager
The DiskIO|Sync and DiskIO|Async Container metrics will not be shown in vRealize Operations Manager for Prometheus monitored containers.
Workaround: None
The Events metric under the namespace object is not shown in vRealize Operations Manager
This issue occurs due to the huge response of 100 MB that is received for events api call against the namespaces.
Workaround: None
Adapter instance is in the warning state if customer upgrades Kubernetes MP from 1.5.1/1.5.2 to 1.6
Due to refresh token expiry in TKGI setup, a popup with error message, "Error in accessing the base url of Kubernetes master - /api/v1", is displayed.
Edit the adapter instance which is in the warning state.
Perform Validate Connection.
Accept the certificates:
TKGI (for Kubernetes clusters deployed on TKGI)
Accept the Kubernetes Cluster
Click Ok when the popup error message is displayed. Click Save.
Click Yes on the confirmation popup message window.
Wait for 2 - 3 collection cycles until the adapter instance is in data receiving state.
Basic Auth does not work on Kubernetes MP for Kubernetes 1.19 and above version
Kubernetes deprecates the support of Basic authentication model from Kubernetes 1.19 onwards.
Use Certificate /Token auth to configure adapter instance for Kubernetes 1.19 and above versions.
Deprecation of cAdvisor
cAdvisor integration is deprecated from the latest version of TKG.
Deploy cAdvisor as Daemon or use Prometheus as a collector service with supported exporters.
Time-out error while performing test connection on selecting cAdvisor collector service
Kubernetes Cluster nodes are not routable for performance metrics to be collected from vROps using cAdvisor.
Use Prometheus as a collector service with supported exporters.
Performance metrics may not be collected for container runtimes except for docker
Few performance metrics may not be collected using cAdvisor collector service for container runtimes, except for docker.
Use Prometheus as a collector service with supported exporters.
Default alerts from Kubernetes Management Pack may not work
Default alerts from Kubernetes Management Pack may not work when Prometheus is used as a collector service.
Workaround: None
Localization cannot be found for Prometheus metrics
Due to the dynamic nature of Prometheus, localization cannot be performed for the Prometheus metrics.
Workaround: None
