AWS ParallelCluster 3.6 is now generally available. Key new features include support for automatic health checks for GPU instances and support for Red Hat Enterprise Linux (RHEL8). Other important features in this release include:

Ability to customize Slurm settings not managed by ParallelCluster

A programmatic interface to manage ParallelCluster using AWS CloudFormation

Support for up to 50 queues and a total of 50 compute resources per cluster

Tag-based cost monitoring in ParallelCluster UI

Support for custom resource tags for queues, head node, and ParallelCluster-managed storage

Extended Amazon CloudWatch metrics for disk usage, idle instances, and errors

Improved head node resiliency with configurable log rotation