Hello, We have a problem where changed default accounts is not taking effect unless slurmctld is restarted.
Example:
My user is a member of two associations, and have matter1 as the
default account:
[root@muon ~]# sacctmgr list users withassoc cluster=matter where
user=paran format=ID,User,Account,DefaultAccount
ID User Account Def Acct
------ ---------- ---------- ----------
1634 paran matter1 matter1
1632 paran nsc matter1
A submitted job correctly gets the account matter1:
[paran@matter ~]$ sbatch -n1 -t 10 shortjob.sh
Submitted batch job 37093
[paran@matter ~]$ squeue -o %a -j 37093 -t all
ACCOUNT
matter1
I then change my default account to nsc:
[root@muon ~]# sacctmgr -i update users set defaultaccount=nsc where
user=paran,cluster=matter
Modified users...
paran
sacctmgr is updated:
[root@muon ~]# sacctmgr list users withassoc cluster=matter where
user=paran format=ID,User,Account,DefaultAccount ID User Account
Def Acct
------ ---------- ---------- ----------
1634 paran matter1 nsc
1632 paran nsc nsc
In slurmctld.log I see an ACCOUNTING_UPDATE_MSG RPC having been
processed:
[2011-02-25T12:25:39] debug2: Processing RPC: ACCOUNTING_UPDATE_MSG from
uid=103
[2011-02-25T12:25:39] debug: updating assoc 1632
[2011-02-25T12:25:39] debug2: association rec id : 1632
[2011-02-25T12:25:39] debug2: acct : nsc
[2011-02-25T12:25:39] debug2: cluster : matter
[2011-02-25T12:25:39] debug2: RawShares : 1
[2011-02-25T12:25:39] debug2: NormalizedShares : 0.020408
[2011-02-25T12:25:39] debug2: LevelShares : 7
[2011-02-25T12:25:39] debug2: GrpCPUMins : NONE
[2011-02-25T12:25:39] debug2: GrpCPUs : NONE
[2011-02-25T12:25:39] debug2: GrpJobs : NONE
[2011-02-25T12:25:39] debug2: GrpNodes : NONE
[2011-02-25T12:25:39] debug2: GrpSubmitJobs : NONE
[2011-02-25T12:25:39] debug2: GrpWall : NONE
[2011-02-25T12:25:39] debug2: MaxCPUMins : NONE
[2011-02-25T12:25:39] debug2: MaxCPUs : NONE
[2011-02-25T12:25:39] debug2: MaxJobs : NONE
[2011-02-25T12:25:39] debug2: MaxNodes : NONE
[2011-02-25T12:25:39] debug2: MaxSubmitJobs : NONE
[2011-02-25T12:25:39] debug2: MaxWall : NONE
[2011-02-25T12:25:39] debug2: Qos : normal
[2011-02-25T12:25:39] debug2: User : paran(1325)
[2011-02-25T12:25:39] debug2: UsedJobs : 0
[2011-02-25T12:25:39] debug2: RawUsage : 206.984109
[2011-02-25T12:25:39] debug: updating assoc 1634
[2011-02-25T12:25:39] debug2: association rec id : 1634
[2011-02-25T12:25:39] debug2: acct : matter1
[2011-02-25T12:25:39] debug2: cluster : matter
[2011-02-25T12:25:39] debug2: RawShares : 1
[2011-02-25T12:25:39] debug2: NormalizedShares : 0.011905
[2011-02-25T12:25:39] debug2: LevelShares : 12
[2011-02-25T12:25:39] debug2: GrpCPUMins : NONE
[2011-02-25T12:25:39] debug2: GrpCPUs : NONE
[2011-02-25T12:25:39] debug2: GrpJobs : NONE
[2011-02-25T12:25:39] debug2: GrpNodes : NONE
[2011-02-25T12:25:39] debug2: GrpSubmitJobs : NONE
[2011-02-25T12:25:39] debug2: GrpWall : NONE
[2011-02-25T12:25:39] debug2: MaxCPUMins : NONE
[2011-02-25T12:25:39] debug2: MaxCPUs : NONE
[2011-02-25T12:25:39] debug2: MaxJobs : NONE
[2011-02-25T12:25:39] debug2: MaxNodes : NONE
[2011-02-25T12:25:39] debug2: MaxSubmitJobs : NONE
[2011-02-25T12:25:39] debug2: MaxWall : NONE
[2011-02-25T12:25:39] debug2: Qos : normal
[2011-02-25T12:25:39] debug2: User : paran(1325)
[2011-02-25T12:25:39] debug2: UsedJobs : 0
[2011-02-25T12:25:39] debug2: RawUsage : 206.987103
However jobs that paran submits still get the old default
account, matter1:
[paran@matter ~]$ sbatch -n1 -t 10 shortjob.sh
Submitted batch job 37094
[paran@matter ~]$ squeue -o %a -j 37094 -t all
ACCOUNT
matter1
The only way I have found to make the change take effect is to
restart the slurmctld.
The cluster runs SLURM 2.1.15. The SlurmDBD runs SLURM 2.2.0 and is
shared with another cluster. (We are planning on upgrading the cluster
to 2.2 soon.)
Regards,
Pär Andersson
NSC
pgpPi5A9w3Rff.pgp
Description: PGP signature
