Re: [Condor-users] Job Configuration

[Johnson koil Raj <johnson.raj@xxxxxxxxx>]

Hi Senthil,


Natarajan, Senthil wrote:

Hi,

I have condor installed on Quad Core Windows machine, so four jobs are running fine using slots 1 to 4.

 

Instead of running 4 windows jobs, I was trying to run 2 Windows jobs and 2 Linux jobs. So I am running Linux Virtual Machine with 2 CPU to run two Linux jobs.

 

How to configure in such a way that two jobs should run on Windows Host and two jobs should run on Linux Guest OS, but as soon as the windows user try to use windows machine, the jobs on both Windows and Linux OS should be suspended.

 

Right now I have configured VMP_HOST_MACHINE on Virtual machine and VMP_VM_LIST on Host machine, the problem is if no one is using the machine, only one job (either Linux or Windows job) is running, other slots changes to Owner status.

 

Please let me know.

Thanks,

Senthil

 

 

From: Natarajan, Senthil
Sent: Wednesday, March 24, 2010 3:08 PM
To: 'Condor-Users Mail List'
Subject: Job Configuration

 

Hi,

I am having condor on Quad Core (windows OS) machine, so there are total of 4 jobs can run using slot1 to slot4.

 

I installed VMWare server and running Linux Virtual Machine with 2 CPU , so two jobs can run on Linux OS, and configured  windows condor to run 2 jobs (using NUM_CPU=2).

 

And also configured VMP_HOST_MACHINE on Virtual machine and VMP_VM_LIST on Host machine. Now maximum of only one job is running if no one is using Host and Guest OS (other slots status changes to owner). It use to run max of 4 jobs when I was using condor on just Windows Host OS.

If you install Condor both on Host and inside VM and change both HOST & VM, the START _expression_ for Host Startd will be set to "START = FALSE" such that Host Startd will not accept Condor job any more. And START _expression_ for VM Startd will remain unchanged.

That is, When a VM starts, the VM will send a registration packet to Host Startd. The IP address of Host machine need to be specified in Condor config file for VM with "VMP_HOST_MACHINE = host.domain.com"

When a Host Startd receives the registration packet from VM, the Host Startd changes "START = " _expression_ to "START = False" unless the Host has already running Condor Job. Then Host Startd sends an reply of "OK" to Host Startd.

If there is a running Condor job on Host machine when Host Startd receives the registration packet from VM, Host Startds sends an replay of "No" to VM. In that case, "START" _expression_ in Host machine will remain unchanged. And "START" _expression_ in VM will be set to "START = FALSE"

That's why the other slots went into owner state.

https://lists.cs.wisc.edu/archive/condor-users/2008-April/msg00257.shtml

http://www.cs.wisc.edu/condor/manual/v7.5/3_13Setting_Up.html#SECTION004138000000000000000

 

Could you please let me know how to configure START, SUSPEND, CONTINUE, PREEMPT, so that if nobody is using Windows Host and it is idle, then I would like two jobs to run on Windows and two jobs on Linux Virtual machine (i.e. all slots on both Host and Guest OS should run jobs). As soon as someone starts using Windows Host, then all the jobs (2 on Windows and 2 on Linux) should be suspended.

Just give a try to this method.

Configure the Host for dynamic slots and to start VM Jobs also. Then start the Linux VM as a VMJob(with Cpu = 2). Once the VM is up and runing configure to report your condorpool. At this point of time your pool has one Windows and linux host with 2 CPUs each.  so you can start 4 jobs 2 with windows and 2 with Linux. When a user started using the host machine configure to suspend the host so the VM Job will also get suspended. ( Here the VM runs as a independent node. i.e not configuring  VMP_VM_LIST and  VMP_HOST_MACHINE.)

I don't know the exact behaviour when the VM Job is suspended and status of condor and job running inside the VM.

 
 

I have following for START, SUSPEND, CONTINUE, PREEMPT.

 

UWCS_START      = ( (KeyboardIdle > $(StartIdleTime)) \

                    && ( HOST_KeyboardIdle > $(StartIdleTime)) \

                    && ( $(CPUIdle) || \

                         (State != "Unclaimed" && State != "Owner")) \

                    && ( $(HOST_CPUIdle) || \

                         (HOST_State != "Unclaimed" && HOST_State != "Owner")) )

 

UWCS_SUSPEND = ( $(KeyboardBusy) || \

                 ( (CpuBusyTime > 2 * $(MINUTE)) \

                   && $(ActivationTimer) > 90 ) || \

                 $(HOST_KeyboardBusy) || \

                 ( (HOST_CpuBusyTime > 2 * $(MINUTE)) \

                   && $(HOST_ActivationTimer) > 90 ) )

 

UWCS_CONTINUE = ( $(CPUIdle) && ($(ActivityTimer) > 10) \

                  && (KeyboardIdle > $(ContinueIdleTime)) && \

                  $(HOST_CPUIdle) && (HOST_KeyboardIdle > $(ContinueIdleTime)) )

 

UWCS_PREEMPT = ( ((Activity == "Suspended") && \

                  ($(ActivityTimer) > $(MaxSuspendTime))) \

                 || (SUSPEND && (WANT_SUSPEND == False)) )

 

Please let me know

Thanks,

Senthil

Regards,
Johnson.

Please do not print this email unless it is absolutely necessary.

The information contained in this electronic message and any attachments to this message are intended for the exclusive use of the addressee(s) and may contain proprietary, confidential or privileged information. If you are not the intended recipient, you should not disseminate, distribute or copy this e-mail. Please notify the sender immediately and destroy all copies of this message and any attachments.

WARNING: Computer viruses can be transmitted via email. The recipient should check this email and any attachments for the presence of viruses. The company accepts no liability for any damage caused by any virus transmitted by this email.

www.wipro.com