A user case was reported in which a HPC job submitted as part of a workflow failed to register as status="FINISHED" even though the job did complete and data were deposited in the archive. Examining the status log, it appears the 'FINISHED' command that informs xGDB_Procedure.sh to exit the 'wait for data' loop was never received, even though it presumably was submitted by the Agave API. TACC staff confirmed that one of two status messages (the 'FINISH' status) was not sent successfully, and suggested updating the JSON notifications string with code that specifies a persistent submission behavior. This was done by creating a function specifying an array equivalent to the modified JSON notification string in jobs_functions.inc.php, and calling this function in each of the four scripts that launches remote jobs for use in creating the final JSON string sent with each job submission. The scripts work as expected with example data (both standalone and workflow versions) although the persistence feature cannot be verified to work since it seems to be an unusual point of failure (one that I had never seen in all my testing).
A user case was reported in which a HPC job submitted as part of a workflow failed to register as status="FINISHED" even though the job did complete and data were deposited in the archive. Examining the status log, it appears the 'FINISHED' command that informs xGDB_Procedure.sh to exit the 'wait for data' loop was never received, even though it presumably was submitted by the Agave API. TACC staff confirmed that one of two status messages (the 'FINISH' status) was not sent successfully, and suggested updating the JSON notifications string with code that specifies a persistent submission behavior. This was done by creating a function specifying an array equivalent to the modified JSON notification string in jobs_functions.inc.php, and calling this function in each of the four scripts that launches remote jobs for use in creating the final JSON string sent with each job submission. The scripts work as expected with example data (both standalone and workflow versions) although the persistence feature cannot be verified to work since it seems to be an unusual point of failure (one that I had never seen in all my testing).