Proposed Pull Request Change

📄 Document Links

View on GitHub View on Microsoft Learn

--- title: Release Notes v8.5.0 description: Product release notes for Azure CycleCloud public preview v8.5.0 author: adriankjohnson ms.date: 06/21/2024 ms.author: adjohnso --- # CycleCloud version 8.5.0 ## New Features: * CycleCloud supports [Confidential VMs and Trusted Platform Launch](../how-to/vm-security.md) for cluster nodes * CycleCloud supports running [Node Health Checks](../how-to/healthcheck.md) during boot to validate node hardware health prior to joining the cluster with the latest AzureHPC images. See [Azure HPC Health Checks](https://github.com/Azure/azurehpc-health-checks) for details * Cluster creation forms now support: * Configuring disk encryption for all cluster nodes * Assigning a Managed Identity to cluster nodes * Setting a custom Boot/OS Disk size * Enabling Node Health Checks * CycleCloud supports Windows Server 2019 and Windows Server 2022 * CycleCloud now supports HPC Pack 2019 * CycleCloud's Slurm cluster type now enables the Slurm RESTd daemon * CycleCloud's Slurm cluster type now supports Ubuntu 22 * New alternate CycleCloud CLI installer with bundled Python for installation offline or on older systems ## Resolved Issues: * Slurm installation would fail with `KeyError: 'ClusterId'` if tags were changed to lowercase * The Machine Type selection filter for *High performance compute* did not include HBv4 * Traditional CycleCloud CLI installer was not properly checking for Python 3.8+ * CycleCloud service autostart had a race condition with mounting the attached data disks which could cause CycleCloud service start to fail on first boot. * Copying a cluster could lose the associated cluster-init specification * Slurm clusters with spaces in their names would not converge * "No node found for instance" was logged when clusters start * Login page did not show the currently logged-in user * Idle GridEngine nodes were killed immediately after starting a job * Editing a node to make changes then re-importing the cluster would reset those changes * Cluster start/terminate button label would wrap while the cluster was terminating