AWS ParallelCluster 3.16 adds an on-node diagnostics tool

AWS ParallelCluster 3.16 is now generally available, featuring pcluster-diag, a new diagnostics tool integrated into the ParallelCluster AMIs. The tool allows users to run diagnostic checks on any cluster node with a single command, generating structured reports to simplify issue identification. This release also improves cluster lifecycle resilience, including more reliable creation, updates, and image builds. The updated software stack includes refreshed versions of NVIDIA driver, CUDA, EFA installer, and Slurm.

The release enhances cluster stability and operational efficiency by hardening the lifecycle management of HPC clusters. Users can now perform node-level diagnostics without additional tools, reducing troubleshooting time. The updated software components ensure compatibility with current NVIDIA and Slurm versions, supporting AI/ML and HPC workloads. AWS ParallelCluster remains an open-source solution for managing HPC clusters on AWS.

AWS ParallelCluster 3.16 is designed for R&D customers and IT administrators who require scalable HPC clusters on AWS. The tool automatically provisions cloud resources into elastically scaling clusters, enabling large-scale scientific and engineering workloads. ParallelCluster is available at no additional charge in supported AWS Regions, with costs limited to the AWS resources used to run applications. The release notes provide detailed information on new features and improvements.

To begin using pcluster-diag, AWS directs users to the Troubleshooting with pcluster-diag documentation. Additional resources include the ParallelCluster User Guide for launching HPC clusters and installation instructions for the ParallelCluster UI and CLI. The release underscores AWS’s commitment to improving HPC cluster management and diagnostics on its cloud platform.