Release notes for past releases of the Windows Machine Config Operator

Before this update, the hybridOverlay service was not using the trusted CA bundle when connecting to OKD, because the --k8s-cacert option was missing from the service command. Because of this, users could encounter trust issues or failures when the hybridOverlay service attempted to communicate securely with OKD clusters by using custom or internal CAs. With this release, the hybridOverlay service command now includes the --k8s-cacert flag that points to the trusted CA bundle. As a result, the hybridOverlay service uses the trusted CA bundle for secure communication, preventing trust issues and ensuring compatibility with the cluster. (OCPBUGS-65856)

Before this update, the hybridOverlay service was not using the trusted CA bundle when connecting to Kubernetes because the hybridOverlay service command was missing the --k8s-cacert option. As a consequence, users could have encountered trust issues or failures when the hybridOverlay service attempted to communicate securely with Kubernetes clusters using custom or internal CAs. With this release, the hybridOverlay service command now includes the --k8s-cacert flag pointing to the trusted CA bundle. As a result, the hybridOverlay service uses the trusted CA bundle for secure communication, preventing trust issues and ensuring compatibility with the cluster.

Before this update, the WMCO neglected to close SSH connections when finishing node reconciliation. As a consequence, after adding a new Windows node to a cluster, the node SSH server would eventually refuse new connections due to being overwhelmed, causing node management issues. With this release, the WMCO now properly closes SSH connections. As a result, the node SSH servers no longer refuse new connections due to this problem. (OCPBUGS-60775)

Before this update, if an internally used config map needed to be deleted and re-created, a nil error was dereferenced when logging the event. As a consequence, the WMCO pod panicked and restarted. With this release, the error handling logic has been reworked. As a result, the Operator pod no longer panics. (OCPBUGS-60792)

Before this update, during secret reconciliations, secret change data was being added to the logs on each reconciliation loop. As a result, this secret change data was persisting, causing the logs to grow in size with unrelated data. With this release, only the current secret change data is being logged, reducing the size and complexity of the logs. (OCPBUGS-61832)

Before this update, when using the optional_namespaces parameter in an ImageTagMirrorSet CR, Windows nodes could fail to pull the specified image, resulting in a image not found error. With this release, the optional_namespace parameter works as expected. (OCPBUGS-47696)

Before this update, Windows Server 2019 nodes did not have a running an SSH server because of network instability. As a result, you were unable to SSH into that node. With this release, the WMCO installs the SSH server node creation. As a result, you can SSH into the Windows nodes as expected. (OCPBUGS-56131)

Before this update, because an Endpoint_IP variable was not resolving, the Windows Instance Config Daemon (WICD) repeatedly reported an Endpoint_IP error. With this release, retries are added to ensure that the Endpoint_IP is created before continuing. As a result, the error message is no longer reported. (OCPBUGS-1721)

Some Windows 2019 Bring-Your-Own-Host (BYOH) Window instances could enter a non-ready state after upgrading to 4.19. Red Hat has not been able to reproduce the issue outside the testing environment, and advises caution when upgrading. If you experience this situation, restart the non-ready instance. (OCPBUGS-47696)