You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Sweep raw-manifest infrastructure for components that should be HelmReleases #205
#68 asked for two things: factor the HelmRelease defaults and drift detection into shared components (delivered by #204), and sweep the template's remaining raw-manifest and vendored infrastructure for pieces that could be a HelmRelease instead. The sweep was not part of #204, so it is tracked here.
Audience and problem
Operators of platforms built from this template. A component installed from raw or vendored manifests gets none of the HelmRelease protections: no retry on a failed upgrade, no drift detection, and no chart-version bump from Renovate. It also has to be re-vendored by hand for every upgrade.
Hypothesis
Moving each raw-manifest component that has a maintained upstream chart to a HelmRelease gives it the same retry, drift and update behaviour as everything else, with no change in what runs.
Acceptance criteria
A list of every raw-manifest or vendored component in the bases and provider layers, each marked "has a maintained chart" or "stays raw" with the reason.
Each "has a maintained chart" entry is either migrated in its own PR or filed as its own issue.
Both provider renders and both workload validate runs stay green after each migration.
Size
S for the inventory; each migration is its own slice.
🤖 Generated by the Agentic Engineer (interactive Claude Code session)
Verdict: done. Inventory of every raw-manifest or vendored component in the bases and provider layers, against main at 679fe8b. Configuration resources for controllers that already run from a HelmRelease (cluster policies, issuers, the Gateway, the Coroot and FluxInstance custom resources, secrets plumbing) are not installs, so they are out of scope.
Component
Layer
Verdict
Reason
auth-proxy
bases
stays raw
Our own Traefik forward-auth configuration, already scheduled for removal under #8.
cdi
bases
stays raw
Upstream ships operator manifests only; no maintained chart.
kubevirt
bases
stays raw
Upstream ships operator manifests only; packagers that need a chart generate one from those manifests because no supported chart exists.
testkube/custom-resource-definitions
bases
has a maintained chart
The Testkube chart in the same directory already installs these CRDs by default, so they have two owners: #263.
coredns
bases, hetzner
stays raw
Only overrides the distribution-managed CoreDNS ConfigMap; nothing to install.
minio
docker
stays raw
Local-only S3 stand-in; replaced the Bitnami chart on purpose (reason in its manifest), and the MinIO community edition is in maintenance mode with no maintained chart.
kubelet-serving-cert-approver
hetzner
stays raw
Upstream ships manifests only (the one chart found is a third-party 0.0.1). It is, however, fetched from upstream's main branch unpinned, unlike the platform's pinned copy: #264.
Our own Jobs, CronJobs and secret seeding; there is no upstream chart.
The one "has a maintained chart" entry is filed as its own issue (#263), which satisfies the second criterion; the third (renders and validation stay green after each migration) moves with it. #264 is a separate pinning gap the sweep turned up. Both are under #175.
Activity
chore: agent-claim #205 nonce=51ba979da92d59f8cbc167ab2f24a8c6
devantler commentedon Oct 24, 2026
Verdict: done. Inventory of every raw-manifest or vendored component in the bases and provider layers, against
mainat679fe8b. Configuration resources for controllers that already run from a HelmRelease (cluster policies, issuers, the Gateway, the Coroot and FluxInstance custom resources, secrets plumbing) are not installs, so they are out of scope.auth-proxycdikubevirttestkube/custom-resource-definitionscorednsminiokubelet-serving-cert-approvermainbranch unpinned, unlike the platform's pinned copy: #264.cluster-heartbeat,vault-backup,vault-config,vault-seedThe one "has a maintained chart" entry is filed as its own issue (#263), which satisfies the second criterion; the third (renders and validation stay green after each migration) moves with it. #264 is a separate pinning gap the sweep turned up. Both are under #175.