SMEM is a tool originally based in the tool QMEM that was writed to work withSunGrid Engine. SMEM has been an adaption of the QMEM to Slurm.
SMEM show us the current state of the resources available in a Slurm clusterwith the possibility to choose if we want to view the jobs that are allocatingthe resources.
SMEM show us the current state of the resources available in a Slurm cluster withthe possibility to choose which jobs we want to list (we can indicate to the toolthat we want to see only the jobs that belong to some user, partition or computers).
To be able to run SMEM in our cluster, this tool needs to be executed in a computerthat belongs to the Slurm cluster. The tool needs to have access to the slurmcommands sacctmgr, scontrol, sstat, sinfo and sacct.
As this tool is a php script, we also need to have php interpreter installed in thecomputer.
The only modification that is necesary to be done in the computer i to add this linein the sudo file:
ALL ALL=(ALL) NOPASSWD: /usr/bin/sstat
By default, the sstat command only allow us to show information of our jobs. We needto modify this behavior to be able to run stat to collect some information of all thejobs that are in the cluster. This is done by SMEM using the sudo tool, so it is neededto add the previous line in the sudo file.
It is also advisable to have SMEM available in the PATH.
Usage:
smem[[-u] [<user-list>]] [-p <partition-name>] [-w <host-list>] [[-g] [<resources-list>]]
-u [<user-list>] Shows the resources in use by all the jobs currentlyrunning. If the user list is present, shows only therunning jobs that belong to the users list. The resourcelist has to be comma separated list.
-p <partition-list> Shows information only for the given partitions The partitions has to be.a comma separated list.
-g [<resource-list>] Show the usage of the general resources. If the resourcelist is present, shows only the usage of the resourcesthat are in the list. The resource list has to be commaseparated list.
-w <host-list> Show information only for the given host list. The host listhas to be a comma separated list.
-h Shows this help.