CycleCloud version 8.0.2

  • Article

New Features:

  • Node and cluster events can be published to EventGrid
  • Upgraded built-in CentOS image references to 7.7 for cloud-init support
  • VM images requiring a purchase plan are now supported (e.g., RHEL)
  • Logging level and error reporting (to Email and Microsoft Teams) have a new simplified dialog box
  • LSF scheduler configs now include extended GPU syntax.

Resolved Issues:

  • "Subnet too small" errors (and similar) did not cause nodes to show up as Failed (in red)
  • VM agent errors on VM start from deallocation were not handled properly
  • Mitigated CVE-2020-11996 (DDoS attack against Tomcat)
  • In some cases, the "node installation succeeded" event was missed and a node would stay in the "Failed" state (regression in 8.0.1)
  • Spot VMs would show up with regular pricing (regression in 8.0.1)
  • Scalesets with more than 100 VMs had errors on orchestration (regression in 8.0.1)
  • Termination of unavailable/failed nodes can get stuck and fail to delete the VM
  • Ephemeral OS Disk would fail with "ReadOnly caching" error (regression in 8.0.1)
  • Table views in the browser could not be edited (regression in 8.0.1)
  • "Error in task Collecting metrics" was logged if a node had no VM (regression in 8.0.1)