| LLNL-SPECIFIC RELEASE NOTES FOR SLURM VERSION 2.0 |
| 19 February 2009 |
| |
| For processor-scheduled clusters (*not* allocating whole nodes to jobs): |
| Set "DefMemPerCPU" and "MaxMemPerCPU" as appropriate to restrict memory |
| available to a job. Also set "JobAcctGatherType=jobacct_gather/linux" |
| for enforcement (periodic sampling of memory use by the job). You can change |
| said sampling rate from the default (every 30 seconds) by setting the |
| "JobAcctGatherFrequency" option to a different number of seconds in |
| the slurm.conf. |
| |
| For InfiniBand switch systems, set TopologyType=topology/tree in slurm.conf |
| and add switch topology information to a new file called topology.conf. |
| Options used are SwitchName, Switches, and Nodes. The SwitchName is any |
| convenient name for bookkeeping purposes only. For example: |
| # Switch Topology Information |
| SwitchName=s0 Nodes=tux[0-11] |
| SwitchName=s1 Nodes=tux[12-23] |
| SwitchName=s2 Nodes=tux[24-35] |
| SwitchName=s3 Switches=s[0-2] |
| |
| Remove the "preserve-env.so" SPANK plugin. The functionality is now |
| directly in SLURM. |
| |
| SLURM version 2.0 must use a database daemon (slurmdbd) at version 2.0 |
| or higher. While we are testing version 2.0, set "AccountingStoragePort=????". |
| Once we upgrade the production slurmdbd to version 2.0, this change will |
| not be required. You can likewise test 1.3.7+ clusters with the same port |
| since 2.0 slurmdbd will talk to 1.3.7+ SLURM. |
| |
| SLURM state files in version 2.0 are different from those of version 1.3. |
| After installing SLURM version 2.0, plan to restart without preserving |
| jobs or other state information. While SLURM version 1.3 is still running, |
| cancel all pending and running jobs (e.g. |
| "scancel --state=pending; scancel --state=running"). Then stop and restart |
| daemons with the "-c" option or use "/etc/init.d/slurm startclean". |
