gcp-benchmark-triage

To triage detached GCP benchmark runs, you interpret DONE, FAILED, or STOPPING markers in the logs and validate artifact integrity. This workflow uses gcp_last_run.json to determine the run state and verify if expected outputs are present.

The best way to retrieve missing vm.log and exit_code.txt artifacts is to fetch them directly from run directories using gcloud storage commands. This allows you to inspect vm.log contents and validate artifact availability for diagnosis.

When a GCP benchmark run shows DONE but artifacts disappear, the triage process checks for missing vm.log or exit_code.txt files to detect VM or artifact loss. It validates stored run artifacts to confirm whether partial results exist.

Yes, you need gcp_last_run.json to diagnose a failed GCP benchmark run. It provides the necessary run metadata to start the triage workflow, locate stored run artifacts, and execute gcloud storage fetch commands.

To clean up VM resources after a GCP benchmark run fails or completes, you follow resource cleanup guidance to perform VM cleanup steps and quota checks. This includes managing cache uploads and preflight checks to reduce downtime.