Skip to content

Refresh content for Kubernetes v1.37 (Garhwal) - #12

Merged
sanketsudake merged 4 commits into
mainfrom
refresh/k8s-v1.37
Sep 4, 2026
Merged

Refresh content for Kubernetes v1.37 (Garhwal)#12
sanketsudake merged 4 commits into
mainfrom
refresh/k8s-v1.37

Conversation

@sanketsudake

@sanketsudake sanketsudake commented Sep 4, 2026

Copy link
Copy Markdown
Owner

What

Moves the book's baseline from v1.36 to v1.37 (released 2026-08-26).
Every change is a swap or a clause added where a chapter already made the claim;
no new flows, figures, or diagram edits.

Sources

Only kubernetes.io release and feature blogs, the v1.37 CHANGELOG, and the feature-gates reference.
Third-party summaries were rejected where they disagreed with the CHANGELOG.

Highlights

  • Pod certificates and ClusterTrustBundles (GA) join the identity paragraph in Chapter 2 and the mesh answer in Chapter 12.
  • StorageVersionMigration (GA) replaces "a controller or batch job" in Flow 30.
  • metrics.k8s.io v1, HPA scale to zero (beta, on by default), and per-HPA tolerance (GA) land in Flow 16 and Q 5.7.
  • Memory QoS (beta, on by default), the static-pod Secret/ConfigMap rule, and failCgroupV1 land in Chapter 4.
  • IPVS deprecation and the planned nftables default rewrite the proxy-mode section and Q 7.5.
  • SELinuxMount (GA) gets a concept bullet and a Flow 25 failure bullet.
  • The DRA extended-resource bridge, device taints, and .status.devices (all GA) extend Chapter 9.
  • In-tree gang scheduling (beta, gate off by default) updates Q 3.8, Chapter 11, Chapter 12, and Appendix D.
  • Q 10.7 now upgrades 1.36 to 1.37 and cites node declared features.

Corrections to existing text

  • Server-side sharded list/watch is alpha (v1.36 and v1.37) and filters each controller replica's hash range at the source;
    the book called it beta and described it as cache partitioning across API replicas.
  • Container restart rules are beta and on by default since v1.35, not alpha.
  • HPA configurable tolerance is GA, not alpha.

Checks

  • Every chapter is within its STYLE.md word budget (ch02 3499/3500, ch08 2399/2400, ch09 2402/2400).
  • git diff -U0 chapters | grep for diagram syntax prints nothing, so the diagram gate did not need to run.
  • notes/v1.37-refresh-plan.md records the pass; notes/research-notes.md is the updated source of truth.

Build

  • build/build.py: 90 s Puppeteer launch timeout and one retry per diagram.
    Two CI runs on this branch failed only on the first diagram, 30 s into the build,
    which is Puppeteer's default launch timeout on a cold runner.
    The run after this change passed with no retry needed.

Record the v1.37 (Garhwal) facts that the chapters cite,
correct four entries that had drifted
(HPA tolerance is GA, container restart rules are beta,
server-side sharded list/watch is alpha and filters per replica,
IPVS is deprecated),
and keep the refresh plan as the record of the pass.
Pod certificates and ClusterTrustBundles, StorageVersionMigration,
resilient watch-cache init, metrics.k8s.io v1, HPA scale to zero,
Memory QoS, static pod API references, IPVS deprecation,
SELinuxMount, DRA extended-resource bridge and device taints,
in-tree gang scheduling (beta, gated off), node declared features.
Correct the sharded list/watch description in ch02, ch06 and ch10.
All chapters stay within their STYLE.md word budgets;
no diagram changed.
Rename the split ServiceAccounts paragraph lead,
put KYAML and PodCertificateRequest in alphabetical order,
narrow the nftables netlink claim to what the release blog states,
and bring ch09 back under its word budget.
The first mmdc call on a cold CI runner can hit puppeteer's 30 s
launch timeout and fail only the first diagram.
Set a 90 s launch timeout in the puppeteer config,
retry each diagram once,
and print the head of stderr so the real error is visible.
@sanketsudake
sanketsudake merged commit e0d8154 into main Sep 4, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant