[Mauiusers] Problems with understanding Standing Reservations

[James A. Peltier]

Jan Ploski wrote:
> mauiusers-bounces at supercluster.org schrieb am 11/21/2007 08:53:44 PM:
>
>   
>> I am trying to figure out if my understanding of standing reservations 
>> is wrong or why my standing reservations aren't working.  I'm trying to 
>> do workstation harvesting from 7PM until 7AM, but also make them 
>> available to interactive jobs during the day 7AM - 7PM.
>>
>> I thought that by defining the following in maui.cfg
>>
>> SRCFG[interactive] PERIOD=DAY DAYS=MON,TUE,WED,THU,FRI
>> SRCFG[interactive] STARTTIME=7:00:00 ENDTIME=19:00:00
>> SRCFG[interactive] CLASSLIST=interactive HOSTLIST=atlas,parhelia
>>
>> CLASSCFG[interactive] FLAGS=PREEMPTOR
>>     
>
> AFAIK, the class id for CLASSCFG must correspond to the name of a queue 
> defined in TORQUE and there is no magic 'interactive' queue for jobs 
> submitted with -I. In fact, you can combine -I with -q to submit an 
> interactive job to a particular queue (= class in Maui's terminology).
>
>   
>> That any job submitted with qsub -I would run on the nodes listed, but 
>> not run on any other nodes not listed,
>>     
>
> According to my understanding, the above is true for any job submitted 
> with "qsub -q interactive", and the -I option is irrelevant.
>
>   
>> but when I
>>
>> qsub -I -lnodes=atlas,mem=1m,ncpus=1
>>     
>
> You didn't specify -q...
>
>   
>> I get
>>
>> ///
>> checking job 4342
>>
>> State: Idle
>> Creds:  user:username  group:graphics  class:batch  qos:DEFAULT
>>     
>
> ...and your job was submitted to queue 'batch' (your default one?)
>
>   
>> WallTime: 00:00:00 of 99:23:59:59
>> SubmitTime: Wed Nov 21 11:31:50
>>    (Time Queued  Total: 00:00:32  Eligible: 00:00:32)
>>
>> Total Tasks: 1
>>
>> Req[0]  TaskCount: 1  Partition: ALL
>> Network: [NONE]  Memory >= 0  Disk >= 0  Swap >= 0
>> Opsys: [NONE]  Arch: [NONE]  Features: [NONE]
>> Dedicated Resources Per Task: PROCS: 1  MEM: 1M
>> NodeCount: 1
>>
>>
>> IWD: [NONE]  Executable:  [NONE]
>> Bypass: 0  StartCount: 0
>> PartitionMask: [ALL]
>> Flags:       HOSTLIST
>> HostList:
>>    [atlas:1]
>> PE:  1.00  StartPriority:  1006
>> job cannot run in partition DEFAULT (idle procs do not meet requirements 
>>     
>
>   
>> : 0 of 1 procs found)
>> idle procs: 116  feasible procs:   0
>>
>> Rejection Reasons: [ReserveTime  :    1][HostList     :    5]
>> ///////
>>
>> So it sees that there is a reservation but shouldn't it allow me to run 
>> this interactive job, as the CLASSCFG is set to interactive?
>>     
>
> No, because didn't use the 'interactive' queue. I suspect that you don't 
> even have a queue named 'interactive' in your TORQUE configuration.
>
> Best regards,
> Jan Ploski
>   
I did get it working, however, my understanding of standing reservations 
was in fact off.  As soon as I defined

SRCFG[interactive] RESOURCES=PROCS:2,MEM:2g TASKCOUNT=8

in maui.cfg it began to work as I expected.  It would seem that the 
standing reservation when resources are not defined requires that you 
explicitly require all the resources of the machine, which is not what I 
wanted.  I wanted to allow a portion of the machine to still be 
available to short running jobs.  Adding that allowed any job that 
required 2 or less processors and 2GB or less memory to run.  The 
taskcount allows me to define the total tasks across all 4 of the 
machines that I defined after setting the resources.  It now seems to 
work as it should.  Jobs requiring more than defined resources of the SR 
are queued.