Mailing List Archives

Authenticated access

 		[Computer Systems Lab]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Problems with jobs

I have removed the STARTD daemon from my central manager to try improve
performance. and have also set TESTINGMODE in the config file.
 
Still only 3 vms actually running jobs still.
 
Chris
----- Original Message -----
From: Ian Chesal
Sent: Monday, December 05, 2005 5:03 PM
Subject: RE: [Condor-users] Problems with jobs

Ahh, but do you only have the one schedd? It looks like you have three jobs running (Claimed + Busy) according to your output -- are they all from the same schedd? Maybe there?s another schedd in your system that?s not able to respond to it?s claims in time?

 

- Ian

 

From: Chris Miles [mailto:chrismiles@xxxxxxxxxxxxxxxx]
Sent: December 5, 2005 11:55 AM
To: Ian Chesal
Cc: Condor-Users Mail List
Subject: Re: [Condor-users] Problems with jobs

 

Its a meaty machine. 2gb of memory and 4Ghz CPU (64 bit) its an IBM built cluster

and doesnt have any resource problem that I can see. there isnt much else working on it.

 

Chris

----- Original Message -----

From: Ian Chesal

Sent: Monday, December 05, 2005 4:26 PM

Subject: RE: [Condor-users] Problems with jobs

 

The dreaded Claimed+Idle. It generally happens to us when our schedd can?t keep up with the processing required to start our jobs. Check the resources on your schedd machine: can your machine handle spawing all the necessary shadows? Or is it running out of CPU, memory, disk, etc?

 

- Ian

 

From: Chris Miles [mailto:chrismiles@xxxxxxxxxxxxxxxx]
Sent: December 5, 2005 11:11 AM
To: Condor-Users Mail List; Ian Chesal
Subject: Re: [Condor-users] Problems with jobs

 

Hi Thanks for the response.

 

SUBMIT_SEND_RESCHEDULE has not specified in any of my config files which

means that its automatically set to true does it not?

 

condor_q -ana says jobs being serviced.

 

It seems a lot of machines go into the claimed state but stay idle.

 

tux.neuralgri LINUX       INTEL  Unclaimed  Idle       0.250   512  0+00:34:48
vm1@xxxxxxxxx LINUX       X86_64 Owner      Idle       0.750  2048  0+00:00:02
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.120  2048  0+00:01:20
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:01:21
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.270  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.180  2048  0+00:00:07
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:11
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.050  2048  0+00:00:07
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:10
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.100  2048  0+00:00:07
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:11
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.100  2048  0+00:00:07
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:11
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:07
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.210  2048  0+00:00:11
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.080  2048  0+00:00:08
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:03
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.050  2048  0+00:00:02
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:03
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.160  2048  0+00:00:04
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Owner      Idle       1.000  2048  0+20:15:59
vm2@xxxxxxxxx LINUX       X86_64 Owner      Idle       0.310  2048  0+00:00:02
vm1@xxxxxxxxx LINUX       X86_64 Owner      Idle       1.000  2048  0+20:16:02
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.220  2048  0+00:00:09
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.130  2048  0+00:00:05
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:06
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:04
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.460  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.200  2048  0+00:00:04
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Unclaimed  Idle       0.130  2048  0+00:00:05
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.170  2048  0+00:00:04
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Busy       0.000  2048  0+00:00:06
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Busy       0.020  2048  0+00:00:04
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Busy       0.000  2048  0+00:00:06
vm1@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.110  2048  0+00:00:05
vm2@xxxxxxxxx LINUX       X86_64 Claimed    Idle       0.000  2048  0+00:00:05

 

Chris

----- Original Message -----

From: Ian Chesal

Sent: Monday, December 05, 2005 2:39 PM

Subject: Re: [Condor-users] Problems with jobs

 

 

 

Hi.

 

When im submitting jobs into my pool it seems to take ages to start running the

jobs unless i run condor_reschedule. Is there a way to speed the process up without

me running this command?

 

[Ian Chesal] See: http://www.cs.wisc.edu/condor/manual/v6.7/3_3Configuration.html#11494 -- make sure you have that set to True in the config file on the machine you?re calling condor_submit from. It will automatically issue a reschedule after submission.

 

My second problem is that job results are not returning to me any quicker than If i ran

my jobs one a one machine pool. I.e im checking condor_q and the queue is only going

down 1 at a time at roughly the same speed as if there was only one machine in that pool.

It is also slower than if I actually ran my jobs sequentially on one machine using a batch

file or shell script.

 

[Ian Chesal] What does condor_q -ana say? Are you setting your job requirements such that only one VM in the system is able to match with all your jobs in your cluster? What about the MAX_JOBS_RUNNING setting on your schedd? Make sure that isn?t set to 1.

_______________________________________________
Condor-users mailing list
Condor-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/condor-users