Learning Forum

[ttalukd]

I am facing a troublesome issue with my simulation sweeps. A lot of job errors and engine errors on the cluster. These simulations run completely fine when ran alone. The problem begins when I run a sweep. I am only changing the width of a Si waveguide and plotting some of its modal characteristics. The job is filled with errors.

This is a screenshot of the cluster:

[Lito]

License consumption depends on your resource configuration. Can you send screenshot of the resource config on the cluster. Also you can check the license status and availability on your license server as shown in this post. Each running simulation will require 1 solve license. See this article for more information on the academic license usage and consumption. For the error on your local computer, try to "Restart" the simulation.

[ttalukd]

When we run into multiple job errors when I am running sweeps with a lot of data points (~ 100,000 data points) is there a way to restart all job errors or do I have to right click each job manually and then click restart (what I am doing now unfortunately). For such an expensive software, this seems like a very elementary feature that's missing. If I have around 500 job errors in a 100,000 data point sweep, is there a faster way to re-run the errors? Thanks.

[Lito]

You shouldn't be getting this many job errors when running the sweep in sequence on the local computer. What are the job error details for these simulations not able to run?
In the meantime, try the following:
Use the latest version, Lumerical 2022 R1.3
If you are only running on 'localhost' try to run using 'Intel MPI' or 'Local Computer' as the Job launching preset in Windows. Do not use all the available threads on your local computer.
Lastly, run concurrent parametric sweeps on multiple nodes in your cluster.