There is also a "wckey" (workoad characterization key) available for
accounting
Quoting Michael Gutteridge <[email protected]>:
We use accounts for that purpose:
http://slurm.schedmd.com/accounting.html
Scroll down to the section on "database configuration" and read up on
"sacctmgr" and associations.
Then users would submit with "sbatch -A foo ..." to have their work charged
to the account "foo". You'd need to set up the associations between users
and accounts. Regrettably, I don't know too much about that half (only
getting started with doing that inside Slurm), but I believe that's the
approach you'd need to take.
Hope this helps
Michael
On Wed, Jan 22, 2014 at 1:40 PM, William B Hurst <[email protected]>wrote:
Greetings,
I could not find an answer to this question in any of the documentation
resources, nor from any that you have listed, so I just thought I would
ask.
We are running numerous large clusters, so this is an important
question. While each user belongs to a group, they may be submitting
jobs on multiple projects. We would like to be able to track resource
consumption by project name.
Is there a way for me to add a "project" parameter requirement for job
submissions to slurm?
If so, could you describe how to implement it?
Best Regards,
Brad Hurst
ps: I have included one of the system configurations below
Configuration data as of 2014-01-22T14:41:34
AccountingStorageBackupHost = (null)
AccountingStorageEnforce = associations,limits,qos
AccountingStorageHost = head.tusker.hcc.unl.edu
AccountingStorageLoc = N/A
AccountingStoragePort = 6819
AccountingStorageType = accounting_storage/slurmdbd
AccountingStorageUser = N/A
AccountingStoreJobComment = YES
AcctGatherEnergyType = acct_gather_energy/none
AcctGatherFilesystemType = acct_gather_filesystem/none
AcctGatherInfinibandType = acct_gather_infiniband/none
AcctGatherNodeFreq = 0 sec
AcctGatherProfileType = acct_gather_profile/none
AuthType = auth/munge
BackupAddr = (null)
BackupController = (null)
BatchStartTimeout = 10 sec
BOOT_TIME = 2014-01-14T11:02:03
CacheGroups = 0
CheckpointType = checkpoint/none
ClusterName = tusker
CompleteWait = 0 sec
ControlAddr = head
ControlMachine = head
CryptoType = crypto/munge
DebugFlags = (null)
DefMemPerCPU = 1024
DisableRootJobs = NO
DynAllocPort = 0
EnforcePartLimits = YES
Epilog = /bin/true
EpilogMsgTime = 2000 usec
EpilogSlurmctld = (null)
ExtSensorsType = ext_sensors/none
ExtSensorsFreq = 0 sec
FastSchedule = 1
FirstJobId = 1
GetEnvTimeout = 2 sec
GresTypes = gpu
GroupUpdateForce = 0
GroupUpdateTime = 600 sec
--
Hey! Somebody punched the foley guy!
- Crow, MST3K ep. 508
