feature: add get_schedule for asset schedules

[Flix6x]

Description

• Single endpoint to fetch the schedules of multiple devices as a single schedule.
• We could still decide to also return a list of scheduling jobs, but when we move to a single job (for running a single scheduler for multiple sensors instead of multiple jobs, each running a scheduler for one sensor) such a list wouldn't exist anymore.

Look & Feel

{
    "schedule": [
        {
            "sensor": 1,
            "values": [
                2.15,
                3,
                2
            ],
            "start": "2015-06-02T10:00:00+00:00",
            "duration": "PT45M",
            "unit": "MW"
        },
        {
            "sensor": 2,
            "values": [
                2.15,
                3,
                2
            ],
            "start": "2015-06-02T10:00:00+00:00",
            "duration": "PT45M",
            "unit": "MW"
        }
    ]
}

How to test

• PR includes test that covers new endpoint.

Further Improvements

• Add test for 2 sensors each with a different resolution.
• Requested duration should match all sensor resolutions.

Related Items

Follow-up to #1065.

[victorgarcia98]

Awesome!

Basically pointing some low effort improvements for the test.

[victorgarcia98]

Perhaps we can try to get the schedules right before letting the worker take the jobs (Line 85). Also, before the jobs are enqueued to test the error of a non-existent job.

[victorgarcia98]

Nice!

I wonder if we should reflect that the output of the schedules used a sequential approach..

[victorgarcia98]

Should this one include the scheduler_info as well? That way we can detect which schedules were fallback and which not.

[victorgarcia98]

It looks that there's is quite a large overlap between this logic and the one found in the sensor get_schedule endpoint. I'd say to keep it like that for now given that this endpoint might still get some changes when we introduce the non-sequential algorithm.

[victorgarcia98]

On the side, I wonder about the case where we do a centralized planning and we want the assets to get their schedules. Maybe in this case the devices can query the schedule "blindly" with certain frequency.

[Flix6x]

On the side, I wonder about the case where we do a centralized planning and we want the assets to get their schedules. Maybe in this case the devices can query the schedule "blindly" with certain frequency.

If we do a centralized planning, we plan them all at once, so either none of them are done or all of them are done, right?

[victorgarcia98]

On the side, I wonder about the case where we do a centralized planning and we want the assets to get their schedules. Maybe in this case the devices can query the schedule "blindly" with certain frequency.

If we do a centralized planning, we plan them all at once, so either none of them are done or all of them are done, right?

Yes but the planning needs to happen from only one point. I'm thinking in the case where we do trigger the plan on our side (for example via a CLI on a cronjob) and the devices request their schedules. This is one idea we had for TUNES.

[Flix6x]

Oh yes, you are absolutely right. Thanks for jolting my memory. It should be made possible for the individual devices to get their own schedule using the job ID of the collective scheduling job. I can make that happen, probably by adding a few lines on the original get_schedule endpoint.