Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
This release provides a revamped autoscale API, the ability to specify custom images in the UI, Managed Identity support and a large number of new features and resolved issues.
New Features:
- CycleCloud includes a new flexible autoscaling API for custom scheduler requirements
- Embedded universal Linux jetpack package will execute on any 64-bit Linux platform.
- Added Ubuntu 18.04 support for select cluster types.
- The cluster creation wizard accepts marketplace, gallery, or custom-image resource IDs in addition to the standard list.
- The
ImageNamenode attribute can select an image via marketplace, gallery, or custom-image resource IDs. - The installation wizard prompts for the initial user's SSH public key
- CycleCloud can use the Azure Managed Identity of the VM it is installed on.
- Nodes support a
ShutdownPolicyattribute to determine whether they are terminated or deallocated on shutdown - The clusters page offers a new combined Shut down option that will terminate or deallocate nodes appropriately
- The Remove Nodes dialog box supports a Force option for nodes that are stuck
- The accounts page now shows available VM sizes for your subscription
- Replaced embedded Ganglia packaging with EPEL sourced packages for Enterprise Linux system monitoring.
- The RAID array chunk size can now be user defined
- Nodes authenticate to the cluster's storage account via SAS tokens instead of access key
- Added
HtmlTemplatecluster parameter type which supports dynamic HTML
Resolved Issues:
- Switching credentials on a cluster would sometimes cause the image selector to fail
- Changing a cluster to use a new account credential would cause node startup failures
- The new HB and HC VM sizes did not appear in the "High Performance Compute" category
- The machine-type selector on a cluster showed VM sizes that were unavailable in the subscription
- Price monitor failures could cause VM type selector to not populate.
- The Add Nodes dialog box allowed creating more nodes than the limit
- Cluster-init project upload did not efficiently handle very large files
- Nodes that were waiting on a return proxy could not be deallocated
- Return-proxy tunnels were not disabled when the node was deallocated, leading to spurious warnings
- The button on large tables to show all rows did not display properly
- The provider account page would sometimes cause browser tabs to freeze
- Let's Encrypt certificates were not extracted properly if using native connectors.