We have an MS HPC Pack 2019 cluster, on-premises, in which the users submit their jobs using a plethora of ways. Sometimes using the HPC Job Manager, sometimes using assemblies invoked from Matlab code, and sometimes using custom tools developed in C#.
I am having a look at the Microsoft.Hpc.Scheduler namespace, and several examples found in the documentation. I am trying to pinpoint some instabilities we are suffering in the cluster, if they can be due to the client tools.
We are in a situation in which the scheduler instance is created once:
private static IScheduler scheduler = new Scheduler();
but the connection is called several times by the clients within the same session:
The user, for each job he/she wants to submit, a) connects b) creates and sets job objects, and c) submits. a, b, and c can be repeated multiple times.
Does this pose a problem in terms of resources from the server not freed?
Also, I could not find a trace of a Disconnect method, and the Close() method is not documented, so I would not know what are the actions and whether it should be called by the user or not. Is there any method to call in order to free resources from the connect method?
Thanks for your help.