That solved the problem.
When you submit the job add a "Memory" requirement... or else condor will create one for you based on ImageSize. The automatically generated one is often OK.... but sometimes is a problem.
Here's how I run one simple job where the image size is overestimated for some reason (when actually running, the job never takes up much):
CONDOR_REQUIREMENTS="Memory>2000" condor_run foo
On Wed, Dec 15, 2010 at 5:04 AM, Sónia Liléo <sonia.lileo@xxxxx> wrote:
Some of the jobs I am running get into the idle state.
The condor_q –better-analyze command says that this is because my machines are rejected by my job’s requirements.
However the only requirements that I define in the sub file are the following
requirements = (OpSys == "LINUX" && Arch == "X86_64") and should_transfer_files = YES
But these conditions are fulfilled by some of my machines. So this is not the problem.
Condor_q –better-analyze says that:
The Requirements _expression_ for your job is:
( ( target.OpSys == "LINUX" && target.Arch == "X86_64" ) ) &&
( target.Disk >= DiskUsage ) && ( ( ( target.Memory * 1024 ) >= ImageSize ) &&
( ( RequestMemory * 1024 ) >= ImageSize ) ) && ( target.HasFileTransfer )
And also that the condition that should be removed is,
1 ( ( ( 1024 * target.Memory ) >= 2750000 ) && ( ( 1024 * ceiling(ifThenElse(J
obVMMemory isnt undefined,JobVMMemory,2.685546875000000E+003)) ) >= 2750000 ) )
My questions are:
Where is this condition defined?
I have at the moment 6 GB memory for 4 cores and RESERVED_MEMORY is set to 0.
Why is this condition preventing the jobs from running?