Share via

CycleCloud version 8.0.0

  • Article

Compatibility Changes

  • The cyclecloud group is now used to control admin access on linux nodes instead of the cyclecloud-admin group.

New Features

  • Cloud-Init is now available on images that support it.
  • UltraSSD disks are now supported.
  • Ephemeral OS disks are now supported
  • Node-level metrics are now collected from Azure Monitor instead of Ganglia
  • The node status display now shows the specific errors that may be encountered while starting nodes in a scaleset
  • Jetpack is now available in RPM and DEB format for easier Linux installation.
  • The CycleCloud CLI now supports python 3.6+. (No longer supports python 2.)
  • Added support for Univa GridEngine 8.6.7 to the GridEngine project.
  • Messages from nodes in the cluster are now encrypted with HTTPS
  • Jetpack no longer uses the custom-data field to receive configuration data.
  • Improved CycleCloud boot time:
    • Removed maintenance converges to reduce jitter on tightly coupled jobs
    • Disabled Samba by default, use cyclecloud.samba.enabled = true to enable
    • Removed legacy autostart/autostop, in favor of autostart/autostop API
    • Smaller subset of SELinux packages are installed as needed
    • Removed NTP installation
    • Removed installation of fail2ban
    • Removed AWS/GCP specific code
    • Removed dependencies on EPEL
    • Removed cyclecloud MOTD
  • Streamlined CycleCloud application
    • Removed Grid Engine job monitoring and jobs page
    • Removed data page
    • Removed reports page
    • Removed query-based alerts page
    • Removed submitonce and filesync

Resolved Issues

  • Clusters wouldn't start if the owner's name contained an invalid character
  • Windows jetpack command would have extra nonsensical output.
  • Jetpack would erroneously warn about ascii/utf-8 encoding issues.
  • Cluster create, delete and copy buttons would be displayed but non-functional for users without cluster management permissions
  • Initial site setup did not prompt for provider account configuration.
  • jetpack initialize failed on certain Centos 7 images.
  • A Slurm cluster name containing a space would cause the master install to fail.
  • When submitting PBSpro jobs that are dependent on other jobs, users can now refer to dependencies by job numbers instead of the full job ID
  • ED25519 SSH public keys would fail validation in the web UI
  • Links to individual scaleset VMs were broken due to a change in the Azure portal
  • In some error cases, VMs that failed to boot were automatically re-imaged even when that would not help
  • Scalesets with failed VMs caused more polling than necessary
  • Invalid HTTP requests would display a Tomcat stack trace in the response body
  • User settings cookies did not include the HttpOnly flag
  • Mitigated CVE issues 2013-4002, 2017-9050, 2019-5477, 2019-16892, 2019-17531, 2019-20330, 2020-5258, 2020-7598, 2020-8840