Dask backend: autoscaling in osparc-dask-scheduler

[sanderegg]

When an external cluster is created there must be the following infrastructure available:

• a docker swarm is up
• 1 machine running a dask-gateway, and a dask-scheduler
  - X machines running dask-sidecar(s) to run solvers

currently we manually scale the number of machines in that cluster by using the AWS console.
this shall be automated and we could leverage the autoscaling service to this end.

Tasks

Beta Give feedback

1. ♻️Autoscaling: prepare for computational autoscaling #4709
   a:autoscaling +1
   
2. ♻️ Autoscaling: refactor before 1st draft of computational autoscaling (⚠️ devops) #4874
   a:autoscaling +1
   
3. ✨Autoscaling: 1st draft on auto-scaling computational clusters #4711
   a:autoscaling +1
   
4. Deploy autoscaling with dask sidecar #4764
   a:clusters-keeper +1
   
5. Pass instance types as resource, such that autoscaling can start/stop necessary machines #4765
   a:autoscaling +1
   
6. ✨ Dask sidecar: find out on start on which EC2 instance it runs if available #4919
   a:dask-service +1
   
7. ✨ Clusters keeper: add rpc methods to get EC2 instance capabilities #4942
   a:clusters-keeper +1
   
8. ✨Comp-backend override resources with selected hardware #4954
   a:director-v2 +1
   
9. ✨Computational autoscaling: find out which EC2 type is necessary #4975
   a:autoscaling +1
   
10. ✨Clusters keeper: shutdown all workers on cluster close #5010
    a:clusters-keeper +1
    

[sanderegg]

upscaling will be taken care of in a separate entry