[HTCondor-users] dedicated_scheduler failing to deactivate claims.

[Francesco Prelz <Francesco.Prelz@xxxxxxxxxx>]

Hi,

we are seeing an issue with the Condor dedicated scheduler for the
parallel universe leaving active claims behind. The startd on the
execute node says:

02/10/15 12:18:45 condor_write(): Socket closed when trying to write 28 bytes to <192.168.10.100:36614>, fd is 3
02/10/15 12:18:45 Buf::write(): condor_write() failed
02/10/15 12:18:45 Failed to send response ClassAd in deactivate_claim.

when trying to send the response to a DEACTIVATE_CLAIM command issued
by the dedicated scheduler. Meanwhile, the dedicated scheduler already 
closed the command socket, as DedicatedScheduler::deactivateClaim()  - and
DedicatedScheduler::releaseClaim() as well, on the most recent 
stable and development versions - aren't decoding any response.

Was ship&pray an intended feature for these commands or is this a bug ? 
If this is a feature, should we always let the claims live to the end of 
their worklife ?

Thank you for a friendly ear.
Francesco Prelz
INFN - Milan