Upgrade → Job Performance module
General Upgrade Notes
The Job Performance (SUPReMM) XDMoD module should be upgraded at the same time as the main XDMoD
software. The upgrade procedure is documented on the XDMoD upgrade
page. The ingestion and aggregation
aggregate_supremm.sh must be run after the XDMoD software has been
9.5.0 to 10.0.0 Upgrade Notes
- This upgrade includes database schema changes.
This upgrade adds a new dimension to the SUPReMM realm to allow filtering on overall GPU usage and new statistics for GPU-hour weighted average GPU usage and GPU hours. It also adds new statistics for the core-hour weighted average maximum memory usage and node-hour weighted homogeneity of jobs.
The upgrade script
xdmod-upgrade will modify the fact tables in the database
to add columns to store the new GPU information. After the
completes, the ingestion and aggregation script
be run. If the
aggregate_supremm.sh script has not been run then the portal will display
an error message similar to the one below when trying to view SUPReMM realm data:
SQLSTATE[42S22]: Column not found: 1054 Unknown column 'gpu_time_active' in 'where clause'
Existing jobs in the database will not automatically be re-ingested by the upgrade scripts. So jobs that were already loaded into XDMoD will not show information for the new statistics and GPU filters. You can force reingestion of all jobs by resetting the job ingest status:
$ /usr/lib64/xdmod/xdmod-supremm-admin --action truncate --resource [RESOURCE] -d
And then run the ingest and aggregation script: