sogno-platform / dpsim-api Goto Github PK

Basically, we would like to replicate the previous dpsim-service API but replace analysis with simulation.
The goal is to implement this python example of dpsim.

We would like to have a static version of the OpenAPI spec on a GitHub branch that we can then render using redoc for the SOGNO documentation.

VILLASctrl can already manage kubernetes resources: https://git.rwth-aachen.de/acs/public/villas/controller

We should try to understand what parts of VILLASctrl can be reused.

In order to control the simulator (dpsim), we need to send AMQP messages in the JSON format defined here:
https://villas.fein-aachen.org/doc/controller-protocol.html

To begin we need to be able to start and stop a simulation.
Perhaps this will be later expanded to communicate with dpsim during the run.

Currently the load profile data for a simulation request is sent by using multipart/form-data for the /simulation endpoint post.
This is not consistent with using sogno/file-service for the model data and makes the swagger ui more complicated to implement.
We should move to using json for the /simuation endpoint with a file-service id for the load profile data.
This will mean that the form parameters in the swagger UI will be listed correctly instead of being shown as "application/octet-stream"

Add new parameters to the Simulation form:
domain: DomainType
solver: SolverType
timestep: u64
finaltime: u64

The simulation status is currently recorded as a boolean "ready" in the file object. We should instead have cluster level logging enabled so that for each simulation id we can log the simulation's progress and see where it is stuck or has failed.

It would be good to have an OpenAPI description of the DPsim API so that users can easily write or generate a client. okapi could be used for this feature.

• create placeholder for results in file service
• create endpoint for returning results for given job id
• retrieve results from sogno-file-service when available

This is connected to sogno-platform/dpsim-worker#6

It is currently not possible to change the url of any related resource (rabbitmq, redis, fileservice) leading to possible collisions with other services, this should be changed in the future. Also the password for rabbitmq is hardcoded to pikas default which is not a longterm solution. Redis in the meanwhile has no authentication set up at all. Both are not available from outside the k8s cluster so this is not a big issue but might become one if user what to use existing resources.

Currently the /simulations endpoint returns a dummy simulation.
Implement the returning the array of existing endpoints.
As returning all the simulations will be lots of data, just return a summary for each simulation.

The openapi.json file is updated, but the redoc pages are not.

We need to start working with keycloak to authenticate users.

Minimum implementation:

• roles
  • admin
  • member - observer plus can start simulations
  • observer - can query simulations and look at simulation results

Network files should be stored in sogno file server and the simulation request needs to include the fileID.

• translate file ID to URL in S3 storage using sogno file service ID
• request to dpsim API with file ID (zip of CIM files)
• create job description and send via AMQP to be retrieved by dpsim worker

The load profile data attribute should not be ignored when starting a Simulation.

Status should be updated at least to the following states:

• Requested (API has sent message)
• Initialising (Worker has read message)
• In progress (Simulation has commenced execution)
• Complete