Re: [Condor-users] Windows XP at boot: sync problem between VMserver and condor services!?!

[Matthew Farrellee <matt@xxxxxxxxxx>]

Rob wrote:
> 
> Hi,
> 
> I have installed on my Windows XP pool PC:
> 
> VMware Server 1.0.9 (Build 156507)
> ActivePerl 5.10.0 (Build 1005)
> Condor 7.2.4
> 
> When the PC reboots, the Condor services are started, but the connection to the
> VMware server fails. Here's part of the output of the GAHP log file:
> 
>  VM-GAHP initialized with run-mode 0
>  VMGAHP: Initialize Uids: caller=SYSTEM@NT AUTHORITY, job user=SYSTEM@NT AUTHORITY
>  VMGAHP: Command returned non-zero: C:\Perl\bin\PERL.EXE C:\condor\bin\CONDOR~1.PL check
>  VMGAHP:   Error: The system returned an error. Communication with the virtual machine may have been interrupted. 
>  VMGAHP:   Error: The system returned an error. Communication with the virtual machine may have been interrupted
>  VMGAHP:   (ERROR) Can't execute C:\Program Files\VMware\VMware VIX\vmrun
>  VMware script check failed:
>  VMGAHP: ERROR: the vm_type('vmware') cannot be used.
>  condor_vm-gahp.exe (condor_VM_GAHP) pid 1960 EXITING WITH STATUS 0
> 
> 
> The "condor_status -vm" command on the master PC shows that the VMserver is
> not working with condor.
> 
> 
> Now, when I later manually RESTART the condor services on the Windows PC,
> all goes well and "condor_status -vm" has the Windows PC in its list.
> 
> 
> Hence, I conclude that at boot up the condor services start too early;
> the condor services should wait until the VMware server registration has finished.
> 
> The manual intervention/restart is not an fix for me, as the setup has to work at boot up
> on hundreds of library PCs.
> 
> Is there an easy way to get this fixed; is there a gahp configuration variable, which
> delays for a few seconds the gahp initialization with the VMserver?
> 
> Thank you.
> Rob.

Sounds like you've run into a somewhat known problem. It is true that
Condor can start and check for VMware before VMware has a chance to
start itself. I believe a proposed fix for this is to actually let the
startd check for VMware periodically, if it isn't present, instead of
only once at startup.

Best,


matt