Cannot start a Worker on Parallel Server on R2022b

I'm trying to start MATLAB Parallel Server R2022b for testing purposes on a single computer before deploying it on something larger. So, I have the Network License Manager, MATLAB, MATLAB Parallel Server, and MATLAB Job Scheduler all on the same computer and I want this same computer to be a Worker. I previously had this setup in R2022a without issues. However, in my attempt to update to R2022b, I manually edited the MJS Service in Windows to point to the R2022b files and now I can't start the computer as a Worker. I've tried uninstalling and reinstalling R2022a and b several times now with and without MATLAB Parallel Server in them, but whenever I try to add the computer as a worker in the Admin Center or on the command line, I get this non-descriptive error message:

Here is the output when starting from PowerShell:

[Edric Ellis]

Please contact Installation Support, they will be able to help you get things working.

Answer 1

[James]

Since you recently upgraded MATLAB Job Scheduler (MJS) from a previous release, then it is possible the error is happening because MJS is failing to load an old state that was left by the older version of MJS. In order to fix this, you'll need to clear out the old state by stopping and starting the MJS service cleanly. Please note that doing this will delete the entire MJS database, which includes all previous job data and any stored user credentials. If you want to keep that data, you will have to restart the MJS service from R2022a, and save that data from MATLAB for each job. We do not support migration of old job data when updating MJS.

To do this, you can use the "Destroy MJS service..." option in Admin Center, which is in the right-click menu on the hosts listed in the Hosts section. You can then restart the service, and you should not encounter the issue.

Alternatively the command-line equivalent is:

cd <matlabroot>\toolbox\parallel\bin

.\mjs uninstall -clean

.\mjs install -clean

.\mjs start -clean

[James]

This worked, thank you!