This comes from Bug #2429. Problem was a job with one step. The step finished in a CANCELLED state due to exceeded memory and job finished in a COMPLETED state. When querying off the information about the job with sacct -j <jobid>, the output indicates the job was COMPLETED, which gives the impression to the user that "all went fine" with the job, despite its step was CANCELLED due to exceeded memory. It's true that if you query off sacct -j <jobid>.<stepid> it shows CANCELLED, but we should somehow indicate when sacct -j <jobid> if all their steps finished COMPLETED or if any of them failed. Proposals: a) When querying off sacct -j <jobid>, only flag state COMPLETED if job was completed and all its steps also finished in a COMPLETED state. b) Add an extra field to sacct -j <jobid> indicating whether any of its steps failed or not. c) Open for more proposals...