April 2026 Patch 2 Release

This patch improves the reliability of self-hosted Private Cloud Director deployments and VM High Availability (VM HA), and completes live migration support for VMs configured with a virtual TPM (vTPM). It also resolves several issues affecting networking, storage, and the user interface.

If you recently upgraded to the April 2026 release, it is strongly recommended that you apply this patch. It resolves a self-hosted installation issue that can fill local disk and crash the management-plane database cluster within days of install, completes vTPM live migration support, and fixes several VM HA reliability issues.

Upgrade Notes

• If you have VMs configured with a virtual TPM (vTPM), complete the manual prerequisite steps published in the April 2026 Release Upgrade Notes — aligning the swtpm UID/GID across all hosts and resizing existing vTPM VMs to a flavor with hw:tpm_secret_security=deployment — before live migration will work for those VMs.

Key fixes and improvements

• Live migration for VMs configured with a virtual TPM (vTPM) is now fully supported, including maintenance-mode operations, completing the capability introduced in the April 2026 release.
• Self-hosted management-plane nodes no longer accumulate unbounded log growth that can fill local disk and crash the control-plane database cluster within days of a fresh install.
• VMs configured with the hw:numa_mempolicy flavor extra spec are no longer forcibly terminated by the host's out-of-memory killer when the requested NUMA node is overcommitted.
• Decommissioning a host with pcdctl no longer deletes shared Image Library Service files that are still in use by other hosts with the Image Library role.
• VM HA heartbeat traffic between hypervisors can now use a dedicated network interface, preventing a single uplink outage from pausing every VM in a cluster.
• VM HA host-status updates are no longer evaluated against a stale, never-expiring cache, which previously caused some status updates to be dropped.
• VM HA now picks up a host's updated IP address after its host-liveness network interface is changed, instead of continuing to use the stale address.
• Resolved a certificate validation issue that was preventing VM HA's storage-attachment visibility from working in self-hosted deployments.
• New self-hosted deployments no longer intermittently fail to create VMs because the Networking Service's configuration was missing the Compute Service's authentication password.
• The Networking Service's OVN control-plane database pods no longer crash-loop when clustering is enabled, due to increased CPU and memory limits.
• Creating or cloning a VM with a pre-reserved network port now uses that specific port or private IP instead of assigning a new one automatically.
• Volume deletion on VM termination is now configurable per VM, including for powered-off VMs, so attached volumes are no longer left behind after deletion.
• Changing the host-liveness-check network interface on an existing host configuration no longer requires a network label to already be assigned.
• Password-reset emails now link to the correct reset-password page, so users can complete the reset flow without contacting support.
• Granting a user access to a tenant from the UI no longer inadvertently revokes a required service-account role, which previously left Kubernetes cluster creation stuck.
• Host onboarding with pcdctl no longer fails on hosts with restrictive default umask settings, such as CIS-hardened baselines.
• Resolved multiple self-hosted installation and upgrade issues: incorrect network interface selection on multi-NIC management nodes, container image pull failures with custom registry paths, a stalled cluster-creation cleanup job, out-of-memory crashes during backup restore on high-core-count nodes, and Kubernetes management-plane nodes registering by IP address instead of hostname.

Known Issues and Limitations

The known issues and limitations noted in the April 2026 Release also apply to this patch release.