Re: [Condor-users] Jobs stuck in condor_starter

[Dan Bradley <dan@xxxxxxxxxxxx>]

Steffen,

Is there anything logged in the corresponding ShadowLog on the submit 
side?  If not, it may be helpful to add D_SYSCALLS to the SHADOW_DEBUG 
config setting.

--Dan

Steffen Grunewald wrote:

> I've found a job cluster that won't run. Jobs are matched against a slot,
> output and error files are created, but condor_starter never transfers
> control to the real Executable (which is a Perl script).
>
> In the slot's StarterLog, there are those messages every hour:
>
> 8/19 13:40:48 ERROR "Assertion ERROR on (result)" at line 384 in file NTsenders.C
> 8/19 13:40:48 condor_write(): Socket closed when trying to write 168 bytes to <10.100.200.93:60802>, fd is 5
> 8/19 13:40:48 Buf::write(): condor_write() failed
> 8/19 13:40:48 ERROR "Assertion ERROR on (result)" at line 875 in file NTsenders.C
>
> A by-product is that apparently there are more jobs in R state than slots
> available (809 free slots, 814 R jobs)
>
> How to interpret the assert() error?
>
> Condor version 7.0.4
>
> Regards, 
> Steffen
>
>