IMPORTANT ANNOUNCEMENT on upcoming changes to batch scheduling system on ALL clusters

Date Posted: March 6, 2015

CHPC is ready to start the transition from our current Moab/Torque batch scheduling system  to SLURM (Simple Linux Utility for Resource Management). The motivation for this transition is financial:  SLURM is open source whereas the use of Moab has a license cost.  A secondary motivation is to align more closely with the batch scheduling systems used on national resources (such as a number of the systems available via XSEDE) to allow for a more seamless transition for users that use these sites.

The goal is to have this transition completed before the current license expires June 30, 2015. It also makes sense, for usage accounting purposes, to make this transition at the start of a new allocation period.  Therefore we have a two-step transition plan:

1. LONEPEAK: The transition began on Tuesday March 10. As nodes complete current batch jobs they will be moved to SLURM.  We strongly suggest that users take the opportunity to test out SLURM batch scripts before the second stage.
2. ALL OTHER CLUSTERS:  The transition will be made in conjunction with the April 2 downtime. When the clusters are returned to service after the downtime, they will be using SLURM.

The content of the cluster user guides as well as any sample scripts provided in the software documentation will be modified to reflect this change, and we will add documentation on SLURM, including a comparison of commands and environmental variables between our current Moab/Torque world and the SLURM world as well as links to other resources available for more information.

CHPC staff will be watching the transition closely and will be available to provide assistance.  We will also hold training sessions, dates/times to be announced, during this transition window.