AWS PCS adds compute-node scaling logs with opt-in delivery

On September 30, 2026, AWS announced scaling logs for AWS Parallel Computing Service (AWS PCS), its managed service for running and scaling high-performance computing workloads using Slurm. The feature records how PCS scales a cluster’s compute node groups, providing node-level detail for troubleshooting. AWS announcement

What the scaling records reveal

Each log entry records one state transition for one compute node. AWS lists instance launches, node registrations, scale-downs and launch failures with their reasons as examples. According to AWS, operators can use the logs to investigate why a node group missed its target size, identify launches that failed for capacity reasons, and check when a particular node started or stopped. Scaling-log details

For engineers, that granularity matters because a group-size shortfall describes the outcome, while a launch failure or registration event provides a more specific clue. Node-level transitions can help focus an investigation, although recording a capacity failure does not resolve the capacity shortage.

Opt-in delivery across PCS Regions

Scaling log delivery is opt-in. PCS supports delivery to Amazon CloudWatch Logs, Amazon S3 and Amazon Data Firehose. AWS says the feature is available in all AWS Regions where AWS PCS is available. Availability and delivery details

The immediate operational step is to enable delivery to a destination that fits the team’s troubleshooting workflow. Because delivery is optional, teams should verify that it is configured before depending on scaling logs during an incident. Treat the records as an additional diagnostic input alongside scheduler and workload information, rather than as a complete explanation of job performance.