SIG Release · Docs Subteam · Shadow Onboarding

Kubernetes v1.38
Docs Team

Everything you own this cycle, every date that matters, and the exact words you send. Keep this open — sections 03, 06, 07 and 10 are the ones you'll come back to.

Cycle beginsMon 31 Aug 2026
Release dayWed 16 Dec 2026
Length16 weeks
Docs Lead@Caesarsage
01Who's who

The people you'll be working with

Two teams matter to us: the v1.38 release team, and our own six.

v1.38 Release Team

RoleLeadWhat we need from them
Release Team LeadSreeram Venkitesh · @sreeram-venkiteshCarries our deadlines in the kubernetes-dev announcements; final say on exceptions
EnhancementsChengHao Yang · @tico88612Owns the Release Tracking board we live in
CommunicationsArsh Sharma · @RinkiyaKeDadRelease blog and feature blogs — we merge the release blog on release day
Release SignalKeisuke Ishigami · @kei01234keiCI and bug triage state going into Code Freeze
Branch Managernot yet assignedCuts the release. We need the exact cut time on 16 Dec from them
DocsDestiny Erhabor · @CaesarsageThat's me — ping me any time, about anything

Our team

Six of us, across five timezones.

NameSlackTimezone
Destiny Erhabor@Destiny ErhaborWAT (UTC+1) — Docs Lead
Abraham Alfaro Sosa@Abraham AlfaroUTC−6
Anshuman Tripathi@anshumanPT
Chiamaka Chielo@chieloGMT+1
Subhasmita Swain@subhasmitaGMT+2
Yashasvi Misra@Yashasvi MisraIST

You have one buddy for the whole cycle, and you share accountability for both of your KEP slices: either of you can answer an author, update the board, or continue a thread. Buddy assignments are in the responsibility sheet. I'm buddied up too, with a KEP slice of my own — this isn't a hierarchy, it's six people covering ~60 enhancements across every timezone.

02Our remit

What the Docs team owns

Since v1.36, Release Notes folded into Docs. That means we ship two things at once.

Documentation

  • Find every enhancement that needs docs, and track it to merged
  • Keep dev-1.38 healthy and in sync with main
  • Review docs PRs for quality, and get technical accuracy from the owning SIG
  • Merge the docs and unfreeze the website on release day

Release notes

  • Generate notes for every alpha, beta and rc
  • Edit them to the style guide, all cycle, not at the end
  • Collect Known Issues and validate major themes with SIG leads
  • Land the Final Release Notes PR before the release is cut
The mantra is "hurry up and wait." Weeks 1–8 are light: half an hour to two hours a day. Weeks 12–16 are heavy: five-plus hours a week, and release day is a long one. Plan your other commitments around the back half, not the front.
03Calendar

The whole cycle on one page

Sixteen weeks. Two of the three capacity holes fall after Code Freeze, which is why our internal dates run a week ahead of the published ones.

w1w2w3w4w5w6w7w8w9w10w11w12w13w14w15w16
KubeCon China KubeCon NA Thanksgiving Enhancements Freeze Code Freeze v1.38.0 Placeholder PRs Ready for review Docs Freeze alpha.1 alpha.2 alpha.3 beta.0 rc.0 rc.1

Every date twice — theirs and ours

The left date is what the community sees. The right date is what we tell KEP owners and what we hold ourselves to. The gap is buffer we never spend.

PublishedOur target
Enhancements Freeze
officialTue 29 Sep (AoE)
Wed 30 Sep 12:00 UTC
noteKEP list is stable — outreach starts 1 Oct
Placeholder PRs open
officialThu 29 Oct (AoE)
Fri 30 Oct 12:00 UTC
we ask forThu 22 Oct
Code Freeze
officialMon 16 Nov (AoE)
Tue 17 Nov 12:00 UTC
noteBurndown meetings begin Mon 23 Nov
PRs ready for review
officialTue 24 Nov
we ask forTue 17 Nov
Docs Freeze
officialThu 3 Dec (AoE)
Fri 4 Dec 12:00 UTC
we ask forTue 24 Nov — merged
Final Release Notes
officialWed 16 Dec
merged byMon 14 Dec (AoE)
v1.38.0 released
officialWed 16 Dec
noteWebsite merge + blog + unfreeze, ~4 hours
Why AoE? "Anywhere on Earth" means the day ends when the calendar day finishes in the last timezone on Earth, UTC−12. A Thursday AoE deadline gives contributors their full Thursday everywhere, so the real cutoff is Friday 12:00 UTC. Never write "EOD" — it's ambiguous. Write the AoE date and the UTC cutoff, both.
The squeeze to watch. Ready-for-review is Tue 24 Nov and Docs Freeze is Thu 3 Dec (AoE) — nine calendar days, but US Thanksgiving lands on Thu 26 Nov and most SIG Docs approvers are US-based. That's roughly five working days of review capacity for the whole release. Everything we do in weeks 9–12 exists to make that week survivable.
04Your job

What you personally own

Three layers: a standing KEP slice that's yours all cycle, a buddy, and rotations that move week to week.

Your KEP slice — end to end

You'll be assigned roughly 10–14 tracked enhancements, grouped by SIG and by shared owner so one message can cover several. "End to end" means you own all of this, not just the first step:

  • Reminders R1 through R5 (section 06)
  • Reviewing the docs PR when it lands
  • Keeping Docs Status current on the board — the same day you learn something, not at the end of the week
  • Logging every reminder in Sheet 7 of the responsibility sheet
  • Telling me early when a KEP goes quiet. Early is free; late is expensive

Rotations — from the responsibility sheet

RotationCadenceYour share
Release notes — assigneeOne per alpha / beta / rc≥1 each, aim for 2
Release notes — reviewerOne per release notes PR≈2 each
Weekly branch syncEvery Friday3 each
Release team updateWeekly, APAC + EMEABy region
SIG Docs updateAPAC + EMEA instancesBy region
Burndown updateMon/Wed/Fri from 23 Nov, then dailyBy region
By the end of the cycle every one of you should have generated release notes at least once, reviewed them at least once, run at least two branch syncs, and given at least two verbal status updates. That's not busywork — it's the set of things a future Docs Lead needs to have done. Two of you should finish 1.38 able to lead 1.39.

If you can't make something

Let the team know 2–3 hours ahead in your meeting's timezone, especially if you're on the hook for a responsibility that day. Everyone has other priorities and things come up — advance notice just lets us redistribute. No explanation needed.

05Release notes

The release notes rotation

Every tag gets a notes PR: one assignee, one reviewer. Nine rows, two of which don't exist in the handbook.

ReleaseTag datePR merged byWeek
v1.38.0-alpha.1Wed 23 SepFri 25 Sep4
v1.38.0-alpha.2Wed 7 OctFri 9 Oct6
v1.38.0-alpha.3Wed 21 OctFri 23 Oct8
v1.38.0-beta.0Wed 4 NovFri 6 Nov10
interim --fix passno tagFri 13 Nov11
interim --fix passno tagFri 20 Nov12
v1.38.0-rc.0Thu 3 DecMon 7 Dec14
v1.38.0-rc.1Wed 9 DecFri 11 Dec15
Final Release NotesleadMon 14 Dec (AoE)16
Why the two interim passes. There's a four-week gap between beta.0 on 4 Nov and rc.0 on 3 Dec with no tag to cut against. A month of unreviewed notes would land at rc.0, ten days before the final PR has to merge. So we clean the accumulated draft on 13 and 20 Nov even with no tag to generate from.

Running it

  • Install krel and always use the latest build — it gets fixes every cycle.
  • Generate the bundle for the tag, then open a PR against kubernetes/sig-release.
  • Edit with krel release-notes --fix. It's session-based, so you can pause and resume. Edits are saved as map YAML files.
  • Always keep release-notes-draft.md and release-notes-draft.json in sync. The JSON drives tooling; drift between them is silent and painful.
  • Cross-check the note count against merges in k/k since the previous tag — krel can skip PRs during rc fast-forward syncs.

Style rules that cause the most churn

Do

  • Past tense, verb first: Added, Fixed, Promoted, Graduated
  • API kinds bare in PascalCase: ConfigMap, ResourceClaim, EndpointSlice
  • Backtick flags, fields, commands, components, metrics, feature gates, versions
  • Graduation phases in Start case: Alpha, Beta, GA, Stable
  • One change per bullet, capitalised, ending in a period

Don't

  • Backtick an API kind — that blocks possessive apostrophes
  • Backtick or quote a field value: "set imagePullPolicy to Always"
  • All-caps ALPHA / BETA
  • Put the period inside a closing quote — it goes outside
  • Use "we", "just", "simply", or Latin abbreviations

A note has to stand alone: a reader should understand the change without opening the PR. If one is too vague to do that, open the PR and rewrite it with the actual symptom.

06Outreach

The reminder ladder

Five touches to get placeholder PRs open. The channel changes as the problem changes — the first two are informational, the rest are human.

R1board

GitHub comment on the KEP issue

Does this enhancement need docs? An explicit "no docs needed" is as useful to us as a PR. Sent by whoever owns the KEP.

send by Thu 1 Oct
R2board

GitHub follow-up

Chase the silent ones and the ones with no PR. State our internal target, 22 Oct, and why: KubeCon NA and Thanksgiving are both in the review window.

send by Thu 8 Oct
R3slack · pair

Slack DM group — first human nudge

Your pair sends this together. Add your buddy to every DM group. By now the problem isn't that they lack information.

send by Thu 15 Oct · 80% checkpoint
R4slack · pair

Slack follow-up, and escalate by SIG

Still silent after two GitHub comments and a DM? Post them in #chairs-and-techleads, grouped by SIG.

send by Tue 20 Oct
R5both

Final call — GitHub and Slack

Three days before the official deadline. Name the consequence: enhancements that need docs and don't have them may be dropped from the release.

send by Mon 26 Oct · 90% checkpoint
Pair protocol, non-negotiable. Add your buddy to every DM group. Whoever gets a reply updates the board and Sheet 7 the same day. React ✅ on the pair-up thread when your slice's round is fully sent. If you can't find an owner on Slack, don't stall — check the KEP's recent comments for who's actually driving it, try another active contributor, or ask me.

The ready-for-review and Docs Freeze phases use shorter versions of the same ladder — four touches each, same channel logic. Dates are in the responsibility sheet.

07Walkthrough

Sending a Slack reminder, step by step

This is the page to reopen when R3 comes around. Four steps, and the fourth is the one people skip.

Step 1

Work out who to actually reach out to

screenshot — KEP issue with Assignees panel, /sig + cc @… lines, and other subteams' recent comments circled
  • Start with the Assignees panel on the KEP issue.
  • Check the /sig and cc @… lines in the enhancement description.
  • Then read the recent comments from other subteams — Enhancements and Comms. Who they're @-mentioning is the strongest signal for who's currently driving the KEP, which is often not the original author.

Start with Assignees. If they look stale, the people other subteams are @-mentioning are the ones actually working on it.

Step 2

Copy the template from our group chat

screenshot — pinned template message in the Docs Slack group, ⋮ → Copy message menu open
  • The outreach template is pinned in our group. Hover the message, then ⋮ → Copy message (or Ctrl/Cmd+C).
  • Only change the author handles and the KEP links. Leave the dates and wording alone.

Never retype the template. Copying the pinned one means every author gets identical wording and identical dates.

Step 3

New group → your buddy → the authors → send

screenshot — fresh Slack DM group with buddy + KEP authors added, template pasted, author replying "Ack, I will open a placeholder PR"
  • Start a new group DM. Add your pair buddy first — before the authors, so you can't forget.
  • Add the KEP authors and owners you identified in step 1.
  • Paste and send. One message per owner group, listing all of their KEPs — that's why slices are grouped by SIG and shared owner.

Buddy first, then the authors. If you forget the buddy, the thread dies the day you're offline.

Step 4

Close the loop

  • Update Docs Status on the board.
  • Log the date in Sheet 7, under the right R-column.
  • React ✅ on the pair-up thread once your whole slice is sent.

This is what turns outreach into a number we can report at burndown. Skip it and our status reports become guesswork.

08Branch sync

The weekly branch sync

Merge main into dev-1.38 every Friday. This is the task most likely to be quietly skipped, and the one that bites hardest on release day.

Use the script rather than the manual steps:

./branch-sync.sh 1.38 -u <your-github-user>from release-team/role-handbooks/docs/branch-sync-script/

The sequence, by hand

Run it manually your first two or three times. The script is faster, but you need to recognise what a sync gone wrong looks like before you let something else do it for you.

CommandWhy
git fetch upstream mainBoth refs current before you start
git fetch upstream dev-1.38
git checkout -b merged-main-dev-1.38-20260911 upstream/dev-1.38Branch from the target. Date it, or next Friday's branch collides with this one
git merge upstream/mainPulls main in. Stops here if there are conflicts
git log --oneline upstream/dev-1.38..HEADSanity check before pushing — see below
git push origin merged-main-dev-1.38-20260911To your fork

Then open the PR against kubernetes/website — and change the base to dev-1.38.

Two ways to get it wrong

Branching off main

The direction is main → dev-1.38. The dev branch receives. Branch off main instead and your PR proposes replacing the dev branch's contents with main — deleting every in-progress feature doc for 1.38.

Leaving the base as main

GitHub defaults the base to main. Leave it and you've opened a PR merging the whole dev branch into the live site — unreleased feature docs and all. Check the base before you hit create, every time.

The sanity check catches both. git log --oneline upstream/dev-1.38..HEAD should print one merge commit plus whatever came from main. Hundreds of commits, or a diff deleting feature docs, means you started from the wrong base — delete the branch and start again.

The Fridays

Sep 4Sep 11Sep 18Sep 25 Oct 2Oct 9Oct 16Oct 23Oct 30 Nov 6Nov 13Nov 20Wed Nov 25 Dec 4Dec 11Tue Dec 15

Two are moved: Nov 27 is the day after Thanksgiving, so that sync happens Wed 25 Nov. Tue 15 Dec is the final pre-release sync and I'll run that one.

When it conflicts

Conflicts are the normal pain point — a page updated on both main and the dev branch. The dev-branch change is almost always an in-progress feature doc, so preserve it and layer the main update on top where they don't clash. If a conflict isn't obvious, git merge --abort and ask in #release-docs before re-attempting. Verify the branch builds before pushing.

Report it every week. Branch health goes in our status update. One or two missed syncs is a yellow status; three or more is red. Say so — nobody is annoyed by a yellow, everyone is annoyed by a surprise.
09Reviewing

Reviewing a docs PR

SIG Docs has limited capacity, so we review actively. A PR needs both a technical /lgtm and a docs /approve to merge.

Mechanics — check every time

  • Targets dev-1.38
  • Has the v1.38 milestone (apply with /milestone if not)
  • No WIP / Draft commits, and no revert commits
  • The KEP is actually likely to land in v1.38
  • Has a technical /lgtm — if not, add the relevant sig/* labels to help find a reviewer

Content — in this order

  • Correct page content type: Concept, Task, or Tutorial
  • Renders properly: numbered lists, code blocks, note/caution shortcodes
  • Makes sense and is spelled right
  • Follows the style guide's best practices and patterns to avoid
Don't block a /lgtm on style nits. Most PR authors are not writers, and many are non-native English speakers. A missing comma or some passive voice should not hold up accurate content. Spend your review energy on correctness and completeness. If a style issue really bugs you, note it and open a follow-up PR.

For technical accuracy, go to the reviewers and approvers named in the KEP's kep.yaml, or the SIG owner. It's fine — expected, even — not to understand every technical component. Get a technical LGTM when you're unsure.

Need help? #sig-docs, and the weekly PR Wrangler is your best resource.

One trap. PRs against the dev branch can merge any time they meet the bar — but be careful not to merge a feature that might slip to the next release. You can add /hold to mark a PR that shouldn't merge until the matching code change has landed.
10The board

Board vocabulary

We live in the Docs view of the v1.38 Release Tracking board. Two fields matter, and every value means something specific.

Board: v1.38 Release Tracking — Docs view

Docs Required?

  • Need docs — needs docs for v1.38. This is the default for every opted-in KEP until the owner says otherwise.
  • Has Docs — docs exist or are in flight
  • None Required — nothing here changes the website
  • Unknown — we read the KEP and still can't tell
  • PR Ready for Review — authored and ready
  • Complete — merged
  • Removed from release

Docs Status

  • No PR — nothing in k/website yet
  • Draft(PR) — open but not ready; may be an empty commit
  • PR Ready to Review — content ready for SIG and SIG Docs review
  • Complete (Merged PR) — reviewed, approved, merged
  • No Docs Needed
  • Removed from Release

After Code Freeze — one more stamp

A few days after Code Freeze, once code-freeze exceptions settle, every tracked KEP gets marked:

StampMeans
Tracked for Docs FreezeDocs PR reviewed, approved and merged — or clearly on course
At Risk for Docs FreezeDocs PR upcoming and the documentation looks incomplete

The four-box template in the templates folder decides which: PR linked in the issue description, targets dev-1.38, in Ready to Review state, and mergeable by Docs Freeze. All four checked means Tracked.

11Escalation

When a KEP goes quiet

There's a defined path. Work it in order, and start earlier than feels necessary.

  1. Comment on the PR and send a Slack DM to the KEP owner.
  2. Ping the SIG chairs and tech leads in #chairs-and-techleads.
  3. Join the SIG's regular meeting and raise it in person.
  4. Last resort — escalate to the Subproject leads and Release Lead. Reach out at least a week before you need it, not the day you need it.
Missing Docs Freeze isn't the end. Owners can file an exception request — ideally in advance if they know they'll miss, otherwise within 24 hours. Exceptions need at least one extra business day after technical review so SIG Docs OWNERS can approve. Tell people this before the deadline; a filed exception is far easier on everyone than a late scramble.
12Status

Reporting status

We report a colour every week. The most important thing to know: SIG Docs actively wants to hear yellow and red early.

GreenOn track. Placeholder or review percentages at or above target, branch sync current, notes PR merged on time.
YellowSomething needs attention or might miss. Below 80% at a checkpoint, one or two missed branch syncs, a notes PR late or blocked.
RedWill miss without help. Below 60% at a checkpoint, three or more missed syncs, or a SIG we can't get a response from.

The handbook's own wording: "SIG Docs prefers and welcomes a status of yellow or red on anything that needs attention, has the possibility to miss deadlines or target dates, or requires follow-up from feature authors and SIGs that you're not receiving."

What to bring to the meeting

Tracked KEPstotal · need docs · no docs needed · unknown
Docs PRsno PR · draft · ready for review · merged
Placeholder PRscount and % against our internal target and the official date
Branch syncup to date, or last synced, or N commits behind
Release notescurrent tag's PR — open, merged, or blocked
Needs attentionspecific KEPs and specific asks, or "none"

Be specific in the last row. "A few KEPs are slow" gets nothing; "SIG Storage KEP #1234 hasn't replied to three touches, can someone from the release team nudge the chairs" gets action.

13Where to be

Meetings, channels, groups

Slack — join all of these

#release-docs#release-notes#sig-docs#sig-release #sig-docs-maintainers#release-management#release-comms#chairs-and-techleads

Plus our private Docs team group, where the templates are pinned and weekly assignments are posted every Monday.

Google Groups

GroupHow to join
kubernetes-sig-releaseJoin directly via the URL
kubernetes-sig-docsJoin directly via the URL
kubernetes-devJoin directly via the URL
release-teamPR to kubernetes/k8s.io → groups/sig-release/groups.yaml — I'll file this for the team
release-team-shadowsSame PR

Meetings

  • Release team meeting — weekly through week 12. APAC instance begins Wed 9 Sep. If you're assigned the update and can't make it, tell the team 2–3 hours ahead.
  • SIG Docs — APAC and EMEA instances. Both need a Docs voice; if you can't attend, put the update in the meeting notes so the community stays informed.
  • Burndown — Mon/Wed/Fri from Mon 23 Nov, then daily with Tue/Thu over Slack from Mon 30 Nov.
  • Our onboarding call — ran Mon 14 Sep, and was not recorded. This deck is the reference instead; ask in the team group rather than by DM and the answer is there for everyone.

Subscribe to the Kubernetes Release calendar so none of this is a surprise.

14Week 1

Your action items

React ✅ in the Slack thread as you finish each one, and reply in-thread with any PRs or issues you open so we can track them.

  • Read the Release Team onboarding guide end to end
  • Review the Community Guidelines and sign the CLA
  • New members: open a dummy PR in kubernetes/contributor-playground now to trigger the CLA bot, so it isn't blocking you later
  • Apply for org membership in kubernetes and kubernetes-sigs, listing sponsors @sreeram-venkitesh and @fsmunoz or @katcosgrove. Skip if you're already a member of both.
  • Join every Slack channel and Google Group in section 13
  • Subscribe to the Kubernetes Release calendar
  • Skim the docs style guide before your first review
  • Skim the Docs handbook, especially the Release Timeline
  • Claim your rotations in the responsibility sheet
  • Install krel and fork kubernetes/website and kubernetes/sig-release
I know this is a lot at once. Ask about any of it — in the group, in a thread, or by DM. We're an inclusive team: if something is unclear or concerning, or if being a shadow is clashing with your availability, just say so. That's a normal conversation, not a problem.
15Links

Where everything lives

The playbook

Tools

Ours

Training

One last thing. If fixing something in the process would make your life easier, do it and open the PR. Both of the last two Docs Leads shipped tooling and handbook improvements while they were shadows, and it's a large part of why the role is more survivable than it used to be. The handbook is ours to improve.