CycleCloud version 8.2.2

  • Article

New Features:

  • CycleCloud now supports terminate notification for VMs
  • The “node table” on the cluster page includes an Azure view which lists VM information
  • Added customizable throttling to improve cluster's response to Insufficient Capacity errors.
  • Shutting down a group of nodes from the UI requires additional confirmation
  • The Slurm integration supports non-autoscaled Slurm clusters
  • Slurm clusters now support login nodes
  • Slurm clusters now support High-Availability configuration of the slurmctlds
  • The Alma Linux 8.5 HPC image is now an option for clusters
  • The unsupported Log4j 1.x version has been replaced with Reload4j

Resolved Issues:

  • In some usage patterns, CycleCloud would use much more disk space than needed
  • Some fonts were being served from a public CDN, causing missing icons in environments with restricted network access.
  • Nodes in an MPI scaleset could sometimes be stuck in Acquiring instead of showing an error
  • The Activity and Monitoring tabs on the clusters page only appeared for users with administrative rights
  • The Windows Remote Desktop username and password were visible to authenticated users without node login permissions.
  • The CycleCloud CLI would error when using Python version 3.8 or higher
  • Slurm nodes would sometimes get Unknown deallocate or Unknown delete errors
  • CycleCloud would hit an error if a load balancer was attached to a scaleset without a public IP