this issue is related to executing a job using $SLURM_ARRAY_TASK_ID. i encountered that the job defined after the first mpi array job in the blueprint is already started right after the first array iteration (out of 61) is done. so the orchestrator doesn't wait for the other 60 slurm array iterations although the job_depends_on relationship option is used in the second job.
this issue is related to executing a job using $SLURM_ARRAY_TASK_ID. i encountered that the job defined after the first mpi array job in the blueprint is already started right after the first array iteration (out of 61) is done. so the orchestrator doesn't wait for the other 60 slurm array iterations although the job_depends_on relationship option is used in the second job.
examplary blueprint: https://github.com/MSO4SC/resources/blob/master/blueprint/zibaffinity/blueprint.yaml array job
za_mpi_em_job
followed byza_prep_nvt_job
which depends on the former mpi array job but starts too early.