Implement upgrade-relation for control plane nodes

[mateoflorido]

Overview

Introduce upgrade orchestration for control plane nodes.

Rationale

To simplify the upgrade process for charms, this pull request adds orchestration logic specifically for upgrading the control plane nodes, specifically the charm core (k8s snap). It does not cover the worker node upgrade orchestration, which will be addressed in a future pull request.

Changes

• Added the on_upgrade_granted handler to manage the upgrade process for nodes in the cluster.
• Integrated the upgrade instance into the status update handler to reflect the current upgrade status in the charm's status.
• Enhanced the snap and charm modules to support unblocking snap installations when the upgrade logic confirms all conditions are met.

[addyess]

Upgrades are obviously hard. We want to guide the user rather than road-blocking them at every turn. In the end, the juju admin knows best what they want their cluster to do and how to get it to the revision they want. We shouldn't make unlimited hurdles from them to cross. This process should guide them -- not block them at every turn.

I say this as one who has performed these upgrades before and putting too many guardrails up saves the casual user but can impede the person who has accepted the risk and just wants the cluster upgraded

[addyess]

this method _announce_kubernetes_version feels similar to the above get_worker_version in some ways that it's reading the version field from the k8s-cluster or cluster relation.

So is a version mismatch now a waiting situation because an upgrade is in-progress? Is that why there's a NOTE here?

[addyess]

Ahh, i see now that it's because the raising of the reconciler error prevented the upgrade events from engaging. Whew -- we still need a check to let folks know they're running on out-of-spec-version of the applications

For now _announce_kubernetes_version is only run on the lead CP. say you deployed and related a kw 1.35 to a 1.31 cluster. I imagine the 1.35 workers may not join. Should they join? Is the k8s-cp the right place to gripe about it? You're right that we should at least make sure we're not in an upgrade scenario before we raise the reconciler error.

[addyess]

Awesome work here. Thanks mateo!

[eaudetcobello]

Nice work! Left a few comments.

[github-actions]

Test coverage for 7e7e480

coverage-report: install_deps /home/runner/work/k8s-operator/k8s-operator/charms/worker/k8s> python -I -m pip install 'coverage[toml]'
coverage-report: commands[0] /home/runner/work/k8s-operator/k8s-operator/charms/worker/k8s> coverage report
Name                                    Stmts   Miss  Cover
-----------------------------------------------------------
lib/charms/k8s/v0/k8sd_api_manager.py     278     29    90%
src/charm.py                              491    242    51%
src/cloud_integration.py                   80      3    96%
src/config/extra_args.py                   27      1    96%
src/containerd.py                         140     16    89%
src/cos_integration.py                     33     12    64%
src/events/update_status.py                51     10    80%
src/inspector.py                           40      4    90%
src/kube_control.py                        39     31    21%
src/literals.py                             6      0   100%
src/protocols.py                           26      5    81%
src/reschedule.py                          77      4    95%
src/snap.py                               185     26    86%
src/token_distributor.py                  181    109    40%
src/upgrade.py                            105     48    54%
-----------------------------------------------------------
TOTAL                                    1759    540    69%
coverage-report: OK (1.22=setup[1.01]+cmd[0.21] seconds)
congratulations :) (1.27 seconds)

Static code analysis report

Run started:2024-12-04 01:55:45.296556

Test results:
  No issues identified.

Code scanned:
  Total lines of code: 3760
  Total lines skipped (#nosec): 3
  Total potential issues skipped due to specifically being disabled (e.g., #nosec BXXX): 0

Run metrics:
  Total issues (by severity):
  	Undefined: 0
  	Low: 0
  	Medium: 0
  	High: 0
  Total issues (by confidence):
  	Undefined: 0
  	Low: 0
  	Medium: 0
  	High: 0
Files skipped (0):

[eaudetcobello]

Thanks!