Skip to content

[PLT-4667] Rewrite the EKS upgrade guide against the real 0.9.0 flow - #943

Open
iamjanr wants to merge 3 commits into
masterfrom
PLT-4667-docs
Open

[PLT-4667] Rewrite the EKS upgrade guide against the real 0.9.0 flow#943
iamjanr wants to merge 3 commits into
masterfrom
PLT-4667-docs

Conversation

@iamjanr

@iamjanr iamjanr commented Aug 19, 2026

Copy link
Copy Markdown
Collaborator

PLT-4667 Rewrite the EKS upgrade guide against the real 0.9.0 flow

Description

upgrade.adoc (es/) rewritten against a real, live-validated 0.9.0 upgrade run (upgrade-provisioner.py, ecr_pull_through.py), instead of the previous 0.7.X→0.8.X-era description:

  • Corrected version range and container image tag to the current 0.9.X scheme.
  • New prerequisites: mounted AWS credentials, helm_repository already publishing the target cluster-operator chart.
  • Corrected the docker run example (the image has no ENTRYPOINT, python3 upgrade-provisioner.py must be invoked explicitly).
  • New "Interactive prompts" section, k8s_version bump step documented in the main flow, expanded final verification (kubelet version per node, node_image-pinned group behavior).
  • Corrected a false claim: node_image-pinned MachineDeployment groups do not get their Kubernetes version bumped either (same as legacy groups) — verified in cluster-operator's keoscluster_controller.go.
  • New "MachinePool" section: mandatory fixed mp_role_name IAM role (vs. CAPA's per-nodegroup dynamic role), with the exact clusterawsadm procedure for clusters that predate this requirement; manual node-group migration steps (the previous "guided migration" script was replaced, see the companion PLT-4265 PR); known MachinePool limitations table.
  • Reordered sections to match the real order an operator needs them in.

Type of Change

  • Documentation update

Related Pull Requests

  • PLT-4265 (Skind, code): the activate-capa-machinepool-features.py and ENIConfig auto-fix this guide documents.

How Has This Been Tested?

Every command and claim in the rewritten sections was validated live against real test EKS clusters going through the actual 0.7.5→0.9.0 upgrade and MachinePool creation flow this session — not just proofread. Rendered with asciidoctor after each change (no warnings, all xrefs resolve).

Checklist

  • [PR title] Title references a Jira ticket.
  • [PR desc] Summary of changes included.
  • [PR desc] Related PRs listed.
  • [PR labels] Correct labels applied — needs ai-assisted at minimum.
  • [Docs] This is the docs PR.
  • Code follows project style guidelines.
  • Self-review performed.

Additional Notes

en/ translation not touched (repo convention: only es/ is authored directly, translation team ports to English).

…ions

Requires mp_role_name for any cluster with MachinePool groups instead of
relying on CAPA's dynamic per-nodegroup role; documents ownership for both
create_iam values and adds a known-limitations table for MachinePool.
Realigns prerequisites, flags, and the MachineDeployment-to-MachinePool
migration steps with the actual behavior of ecr_pull_through.py and
upgrade-provisioner.py, and documents recovery steps for interrupted
upgrades.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant