Bmcjob controller implementation

[pokearu]

Description

The BMCJob controller is responsible for running the Tasks (BMC actions like power and bootdev) listed on a BMCJob CR.
The controller,

1. Fetches the BMCJob object
2. Resolves the BaseboardManagementRef and establishes a connection (BMCCLient) to the baseboard management.
3. Pauses the BaseboardManagement object reconcile. (Defers unpause)
4. Reconciles the individual tasks in the BMCJob spec.
5. For each Task the controller creates a BMCTask object, with an OwnerReference.
6. Runs the action mentioned in the Task.
7. Updated the Task conditions and status.
8. Updates the Job condition and status.

Why is this needed

BMCJob is useful for performing one time actions that we want on the servers.

How Has This Been Tested?

Created a KinD cluster and tested creating and deleting BMCJob objects.

Checklist:

I have:

• updated the documentation and/or roadmap (if required)
• added unit or e2e tests
• provided instructions on how to upgrade

[pokearu]

This PR is a WIP 🏗️

Also need to rebase based on changes discussed in #5

[codecov-commenter]

Codecov Report

Merging #6 (e85f60d) into main (985d2ca) will decrease coverage by 37.03%.
The diff coverage is 3.10%.

@@             Coverage Diff             @@
##             main       #6       +/-   ##
===========================================
- Coverage   58.49%   21.45%   -37.04%     
===========================================
  Files           4        4               
  Lines         106      289      +183     
===========================================
  Hits           62       62               
- Misses         37      219      +182     
- Partials        7        8        +1     

Impacted Files	Coverage Δ
controllers/bmcjob_controller.go	0.00% <0.00%> (ø)
controllers/bmctask_controller.go	0.00% <0.00%> (ø)
controllers/baseboardmanagement_controller.go	70.45% <66.66%> (-2.49%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 985d2ca...e85f60d. Read the comment docs.

[chrisdoherty4]

Please could you add to the description commentary on condition semantics.

[pokearu]

Based on discussions I have refactored how BMCJob and BMCTask resources are controlled.

bmcjob_controller

1. Runs a BMCJob and also watches BMCTask based on owner reference.
2. Initially sets the Job to Running and creates the first BMCTask
3. The reconcile loop gets called again once the BMCTask updated and based on Task Conditions, the next BMCTask object is created.
4. Once all the BMCJob tasks have condition completed, BMCJob is marked Completed.

bmctask_controller

1. Runs the actions on a BMCTask
2. Establishes a connection to the BMC and executes the specifies action.
3. Updates the status

[chrisdoherty4]

Why do we think Failed is a condition? Wouldn't that just be state on the Status (Success or Failed being the valid values)? In the event its Failed a description field of some sort. This way, consumers can watch for the condition "Completed" to become true before inspecting the actual result.

I don't recall what we said about Failed and Success etc. I'd be OK with allowing this to pass and changing it in a separate PR (or stripping it out of this one if possible). There's a bunch of good stuff here already that we don't want to hold up.

We need to be careful that we're not encoding job state into conditions.

[chrisdoherty4]

I minor comment to fix in a subsequent PR.

[chrisdoherty4]

Rufuio.tinkerbell.org