Hi all,
I am trying to test a setup for a cluster using a parametric sweep job. The job has several tasks, each task is a small "hello world" type executable (it writes some text to a separate file for each task to a shared directory on the network). The
problem I have is that all my attempts are given the "Queued" status with the following message:
The job is pending: Not enough available resources.
Background:
The network topology is Enterprise with 3 nodes:
- Header node (Win Server 2012) - 8 cores
- 2 Compute Nodes (each a virtual machine on a different physical machine - Win Server 2008) - 4 cores each
In Node Management all nodes are Online, Node Health OK.
In the default job template I selected the default resource pool (with with of 100 - so all 16 cores are enabled).
I can successfully run Single-Task jobs with the same task as above (write some text to a file). I can submit and execute this on each of the 3 nodes available. So all nodes should be accessible for jobs. However if I try to modify job resources (set the
number of cores from Auto to a any number larger than 1) this will also fail, with the same message as the the parametric sweep task. It seems that the job scheduler cannot find and allocate the needed resources.
I have tried every possible configuration that I see in the job submission interface and they have all failed. So I am guessing either I am doing something terribly wrong, or there is something wrong with the configuration of my HPC cluster.
Any ideas to make this work are greatly appreciated.
Regards,
Valentin
