Release NotesΒΆ

Scyld ClusterWare Release v11.8.1 is the latest update to ClusterWare 11.

For the most up-to-date product documentation and other helpful information, visit the Penguin Computing Support Portal. In particular, the most recent Scyld ClusterWare documentation can be found on the Penguin Computing Support Portal at https://www.penguincomputing.com/support/documentation. The most recent version will accurately reflect the current state of the Scyld ClusterWare yum repository of RPMs that you are about to install.

Important

Previous releases of ClusterWare 11 included the clusterware-ganesha package that provided bindings to the Scyld nfs-ganesha package. Those bindings have been removed. Penguin Computing will continue to provide the nfs-ganesha rpm packages through the end of 2021. If your site is using NFS-Ganesha, please confirm that you have another source for those packages before that date. If you are not using NFS-Ganesha, then those Scyld nfs-ganesha packages can be removed via yum:

yum remove nfs-ganesha*.scyld.*

Release Notes for Scyld ClusterWare 11:

  • Assorted bug fixes on top of 11.8.0. See Changelog for details. Release notes from 11.8.0 continue below.

  • Add support for clusters built on the ARM architecture. Penguin's testing using Rocky 8 confirms that ClusterWare runs properly on ARM-based head nodes and can boot ARM-based compute nodes. Mixed clusters should be possible in a multi-head cluster with at least one head per architecture used for image modification.

  • Recognize when a Red Hat CoreOS (RHCOS) ISO file is used to create a ClusterWare repo and initially populate the kernel command line of any boot configuration that references that repo. The cmdline field includes references to node attributes that will be replaced with the attribute values at boot time. Other attribute substitutions in the boot configuration cmdline are also supported allowing for command line customization for a node or attribute group.

  • Rewrite the "unknown node" handling to keep a list of MAC addresses requesting DHCP and optionally list those MACs via the scyld-reports unknown tool. The MAC addresses are parsed directly from the ISC DHCP log and stored in individual head node memory. Future changes will allow unknown nodes to be automatically added to a specified naming pool.

  • Significantly improve scyld-modimg --capture performance by removing the compression, thereby allowing for much higher throughput completely offsetting the increase in data size.

  • Fix a variety of bugs including one where DHCP leases were not automatically updated when a node's MAC address was changed, another where deleting an in-use naming pool broke scyld-*ctl command line tools, and decrease the boot time of the default rwram boot style.

See Changelog for a full history of ClusterWare releases, and Known Issues And Workarounds for a summary of notable known current issues.