There are several reasons why the parpool stage may fail validation on your client machine during validation:
1. The client machine and cluster worker machine(s) are unable to communicate with each other.
When opening a parpool, the client machine and machine(s) running the MATLAB workers need to be able to have an open communication with each other. As a result, the machines should be able to communicate with each other's hostname and IP address. To test this, on your client node and each worker node, open up Command Prompt/Terminal and run the following commands:
ping <hostname>
ping <IP address>
where the <hostname> and <IP address> are the values matching those of the machine you are attempting to connect to.
2. A firewall exists between the submission node(s) and the worker machine(s)
If the MATLAB Parallel Cluster cluster is behind a firewall, the firewall will need to be configured to allow parpool to pass validation. For more information on configuring your firewall with MATLAB Parallel Server, reference this article: How do I configure MATLAB Parallel Cluster and Parallel Computing Toolbox to work within a firewall?
3. The client machine is on a different subnet than the worker machine(s).
In some network setups, client machines exist on different subnets than the worker machine(s), with the headnode being the only bridge between the two subnets. For these setups, parpool will always fail, as direct communication between the client machines and worker node(s) cannot be established.
NOTE: Starting in R2019a the following name changes occurred:
- MATLAB Distributed Computing Server was renamed to MATLAB Parallel Server
- mdce_def was renamed to mjs_def
- mdce binary was renamed to mjs
