Re: [HTCondor-users] Help: How to use VM Universe? Thank you

[btdan <btdan@xxxxxxx>]

Thank you for your reply.

I have already modified the BIOS setup of the server in order to support the virtualization. And now I make sure the kvm is running now.

And I also modified the condor_config file to support the vm, and the HasVM parameter is True .

The submit file is as follows:

universe                     = vm

executable                   = kvm_sample_job

log                          = sample.vm.log

vm_type                      = kvm

vm_memory                    = 2048

# should_transfer_files = YES

# when_to_transfer_output = ON_EXIT

# transfer_input_files = /data/linux_OS_ISO/CentOS-6.4-x86_64-bin-DVD1.iso

# /data/kvmswap.img

# vm_cdrom_files=/data/linux_OS_ISO/CentOS-6.4-x86_64-bin-DVD1.iso

# vm_should_transfer_cdrom_files=True

vm_checkpoint=True

vm_disk=/data/kvmswap.img:sda2:w:ext4

# kvm_cdrom_device=/media/cdrom

queue

 

After I condor_submit this file. The result of condor_q -better-analyze id  of this job is:

The Requirements _expression_ for your job reduces to these conditions:

 

         Slots

Step    Matched  Condition

-----  --------  ---------

[1]          48  TARGET.HasVM

[3]          48  TARGET.VM_Type == "kvm"

[5]          48  TARGET.VM_AvailNum > 0

[11]         48  TARGET.FileSystemDomain == MY.FileSystemDomain

[15]         48  TARGET.VM_Memory >= 2048

 

Suggestions:

 

    Condition                         Machines Matched    Suggestion

    ---------                         ----------------    ----------

1   ( TARGET.HasVM )                  48

2   ( TARGET.VM_Type == "kvm" )       48

3   ( TARGET.VM_AvailNum > 0 )        48

4   ( TARGET.FileSystemDomain == "xxxx.com" )48

5   ( TARGET.VM_Memory >= 2048 )      48

6   ( TARGET.Arch == "X86_64" )       56

7   ( TARGET.Disk >= 0 )              56

8   ( TARGET.HasFileTransfer )        56

9   ( TARGET.TotalMemory >= 2048 )    56

10  ( ( MY.CkptArch == target.Arch ) || ( MY.CkptArch is undefined ) )

                                      56

11  ( ( MY.VM_CkptMac is undefined ) || ( TARGET.VM_All_Guest_Macs is undefined ) || ( stringListIMember(MY.VM_CkptMac,TARGET.VM_All_Guest_Macs,",") == false ) )

                                      56

 

However ,I get the message from /var/log/condor/ShadowLog  as follows:

 

05/15/13 20:05:47 Using config source: /etc/condor/condor_config

05/15/13 20:05:47 Using local config sources:

05/15/13 20:05:47    /etc/condor/condor_config.local

05/15/13 20:05:47 DaemonCore: command socket at <111.111.11.220:47405?noUDP>

05/15/13 20:05:47 DaemonCore: private command socket at <111.111.11.220:47405>

05/15/13 20:05:47 Setting maximum accepts per cycle 8.

05/15/13 20:05:47 Initializing a VM shadow for job 21.0

05/15/13 20:05:47 (21.0) (5521): Request to run on slot1@xxxxxxxxxxxxxxx <111.111.11.206:53684> was ACCEPTED

05/15/13 20:05:47 (21.0) (5521): File transfer completed successfully.

05/15/13 20:06:07 (21.0) (5521): Job 21.0 going into Hold state (code 6,0): Error from slot1@xxxxxxxxxxxx: VMGAHP_ERR_JOBCLASSAD_KVM_INVALID_DISK_PARAM

05/15/13 20:06:07 (21.0) (5521): **** condor_shadow (condor_SHADOW) pid 5521 EXITING WITH STATUS 112

 

I google the keyword VMGAHP_ERR_JOBCLASSAD_KVM_INVALID_DISK_PARAM, but I get nothing useful.

The parameter of vm_disk in the submiting file is /data/kvmswap.img, it is created with the following command:

  dd if=/dev/zero of=kvm_swap.img count=1000000

  mkfs.ext4 kvm_swap.img

Can somebody tell me how to solve this problem ?

Or give me an sampe submit file ?

Thanks a lot.

---

btdan

 

From: Brian Candler

Date: 2013-05-14 22:18

To: btdan

CC: HTCondor-Users Mail List

Subject: Re: Re: Re: [HTCondor-users] Help: How to use VM Universe? Thank you

On Tue, May 14, 2013 at 09:42:36PM +0800, btdan wrote:

>         56 are rejected by your job's requirements

...

>        Condition                         Machines Matched    Suggestion

>        ---------                         ----------------    ----------

>    1   ( TARGET.HasVM )                  0                   MODIFY TO

>    false

>    2   ( TARGET.VM_Type == "kvm" )       0                   REMOVE

>    3   ( TARGET.VM_AvailNum > 0 )        0                   REMOVE

>    4   ( TARGET.Disk >= 75000000 )       0                   MODIFY TO

>    1727434

>    5   ( TARGET.VM_Memory >= 2048 )      0                   REMOVE

 

There in black-and-white is why your job isn't running.

 

* You asked it to run only on machines which are advertising HasVM = True.

  None of them match this condition.

* You asked it to run only on machines which have VM_Type == "kvm". None do.

* You asked it to run only on machines where VM_AvailNum > 0. None do.

* You asked it to run on machines with >=75GB of disk space. None do.

  It hints that only 1.7GB is actually free.

* You asked it to run only on machines with VM_Memory >= 2048. None do.

 

You clearly need to make config changes on your *execution* nodes to give

them the capability to run VMs, and to announce this capability in their

MachineAd.

 

>    It seems that the kvm is not running?

 

Of course not. Condor hasn't found any machine in your cluster which is

suitable to run this job, so it hasn't started anywhere.

 

So let's go back to your original mail:

 

"I want to use VM Universe. And follwing the instruction."

 

Which instruction exactly are you follwing?  Give the URL.  I suspect you

have missed out a huge chunk.

 

Regards,

 

Brian.