Integrate KEDA into milner #96

Merged

tepichord merged 6 commits from feat/85-integrate-keda into main

2026-09-25 21:19:58 -07:00

tepichord commented

2026-09-22 21:31:56 -07:00

Owner

Closes #85

Automated PR created by forgejo-done adapter.

Closes #85 Automated PR created by forgejo-done adapter.

tepichord added 4 commits

2026-09-22 21:31:56 -07:00

feat(keda): add KEDA namespace, HelmRelease, HelmRepository, and pinned values 9f89731740

Installs KEDA 2.20.2 as an infrastructure-layer Flux component in a new
keda namespace with restricted PSA enforcement. CRDs ship in
templates/crds/ (not the chart's crds/ dir), so helm-controller's
install/upgrade.crds policy never runs — they ride the release manifest.

The CRD_PROVIDERS row in tests/test_gitops.py is deferred to issue #25
step 10 (annia's ScaledJob), which is the first consumer of these CRDs.

Refs: #85

feat(keda): wire KEDA into infrastructure kustomization 221656aebe

Adds namespaces/keda.yaml in the namespaces block and keda in the
component list after rabbitmq-operator.

Refs: #85

docs(keda): add §17 documentation and OPERATIONS.md entries 6ba13005ce

Covers: what KEDA is for here, namespace PSA level, reading ScaledJob
state, pause/resume via autoscaling.keda.sh/paused annotation, and
debugging the external.metrics.k8s.io API service.

Refs: #85

docs(keda): correct ScaledJob state and metrics-server scope

ci/woodpecker/pr/gates Pipeline was successful

Details

be535ccd80

§17c read the ScaledJob's `ScalingActive` condition, but that is an HPA
condition: a ScaledJob has no HPA. KEDA's ScaledJob conditions are `Ready`,
`Active`, `Fallback` and `Paused` (upstream condition_types.go calls
ConditionHPAActive "ScaledObject-specific (ScaledJob has no HPA)"), so the
command returned nothing. Query `Ready`, which is what says the triggers
resolved and the scale loop started.

§17e then attributed a stalled ScaledJob to `v1beta1.external.metrics.k8s.io`.
That API service is the HPA's external-metrics path for a ScaledObject; the
ScaledJob operator drives its jobs directly from the scaler and never reads it.
Scope the check to ScaledObjects so the section does not send an operator to
the wrong component when annia's ScaledJob is quiet.

tepichord added the

reviewing

label

2026-09-22 21:31:57 -07:00

tepichord added this to the MVP project

2026-09-25 20:33:55 -07:00

tepichord added 2 commits

2026-09-25 20:43:49 -07:00

Merge branch 'main' of ssh://clone.gloom.garden/tepichord/milner into feat/85-integrate-keda 0911918aa0

docs(keda): number the KEDA runbook §18, because IRC took §17

ci/woodpecker/pr/gates Pipeline was successful

Details

ci/woodpecker/pull_request_closed/cleanup Pipeline was successful

Details

6e731610f1

The branch was written when §17 was the next free section — issue #85 says as
much — and the IRC runbook (docs/irc.md, merged as #105) claimed §17 first.
Two documents cannot share an address: §17 appears dozens of times across
tests, manifests and scripts, every one of them meaning docs/irc.md §17x, so
KEDA moves instead of IRC. §18 is the next free number, and the sequence is
not meant to be contiguous — §1, §3, §6 and §8 are all skipped.

OPERATIONS.md is the only file that names KEDA's number, so this is the whole
change: the heading and five subsections in docs/keda.md, and the two rows in
OPERATIONS.md.

Refs: #85

tepichord merged commit d3a5ed016c into main

2026-09-25 21:19:58 -07:00

tepichord deleted branch feat/85-integrate-keda

2026-09-25 21:19:58 -07:00

tepichord referenced this pull request from a commit

2026-09-25 21:19:59 -07:00

Merge pull request 'Integrate KEDA into milner' (#96) from feat/85-integrate-keda into main

tepichord removed the

reviewing

label

2026-09-25 21:34:52 -07:00