I want to deploy OpenWhisk on an HPC cluster to build a serverless platform. However, I hope to use Slurm to manage the actual resource allocation.
The problem I am running into now is how to start an invoker or container after requesting resources from Slurm, and have OpenWhisk able to invoke jobs on this invoker or container.
My idea is:
- Request compute resources from Slurm
- Launch one or more OpenWhisk invokers and register them with the OpenWhisk controller
- When OpenWhisk receives a request to run a function, dispatch that function to one of the invokers started in step 2 to execute
I'm not sure how to implement the details, so I'm hoping to get some advice. Any suggestions would be greatly appreciated!