Public: anyone can read this; the crew can change it.

Spin down the Kubernetes platform: everything lives in the forge #7667

shipped production is serving it opened this has it
opened this

Owner, 2026-11-01: "i want to spin down the kubernetes open-platform and have everything in the forge. eventually all of my github and vercel will get migrated as well." Not urgent: shipping the roadmap comes first. Filed so the direction is on record, and so new work does not deepen the dependency.

Outside agents show the cost today. grokbot's work is on git.open-platform.sh (Forgejo, trevato-apps/forge pulls such as #42 to #67, the latest updated 2026-10-26), and none of that friction is visible in this forge.

Done when:

  • [x] An inventory on this issue lists everything the forge and its apps still take from the platform: the image build and registry, gitops and reconcile, the pod and its supervisor, ingress and TLS, backups, other hosted apps, identity. Each has what replaces it. #1432 (self-deploy) is closed, so state what it actually retired and what still goes through Forgejo.
  • [ ] Outside agents work against the forge and are pointed there. The trevato-apps/forge route closes, once #6359 gives them a way in.
  • [ ] Each dependency is cut one at a time with a tested rollback. The forge is never dark without the owner's go.
  • [ ] The platform spins down only after a set period with nothing reaching it, measured from its access logs, and after a restore drill that does not touch it.

GitHub and Vercel go through #6007. Split the work into small issues from the inventory; do not start on cuts before it exists.

commented

Live triage (automatic, not written by a person): no pillar named, so it waits behind the Later gate; name a pillar to move it.

commented

The owner, 2026-11-02: "this does go hand in hand with wanting everything to be forge native now and getting off of the platform and the forge to forge origin connections". So this moves up: start now with the inventory only, in a pull.

For each thing the forge and its apps still take from the platform, list what replaces it, what is already done, and file each next step as its own issue in order. Include: image builds and the registry, gitops and reconcile, pod and supervisor (keeper #8374), ingress and TLS, identity, the backup lanes (forgejo-dump is suspended, 36 GB), the Forgejo mirror where outside people still send pulls (#8935), the hosted apps, and the droplet release tick.

Nothing is turned off or removed by this work. Each cutover is done with the owner around, and the forge never goes dark.

the crew's run was interrupted: restart, resumes on the next pass — cancelled
commented

Reviewed pull #8953 at its head: ticked An inventory on this issue lists everything the forge and its apps still take from the platform: the image build and registry, gitops and reconcile, the pod and its supervisor, ingress and TLS, backups, other hosted apps, identity. Each has what replaces it. #1432 (self-deploy) is closed, so state what it actually retired and what still goes through Forgejo.. left open: Outside agents work against the forge and are pointed there. The trevato-apps/forge route closes, once #6359 gives them a way in. — Only described as a future step (#8950); the route is still open and #6359 has not landed.; Each dependency is cut one at a time with a tested rollback. The forge is never dark without the owner's go. — No dependency has been cut; the page plans the cuts and no rollback has been tested.; The platform spins down only after a set period with nothing reaching it, measured from its access logs, and after a restore drill that does not touch it. — The gate is written down (#8952), but no quiet period has run and no drill has passed..

commented

Left open after pull #8953 landed: 3 unticked items remain (Outside agents work against the forge and are pointed there. The trevato-apps/forge route closes, once #6359 gives them a way in.; Each dependency is cut one at a time with a tested rollback. The forge is never dark without the owner's go.; The platform spins down only after a set period with nothing reaching it, measured from its access logs, and after a restore drill that does not touch it.). The planner files each as its own issue and closes this one; a person may close it, tick what is done, or say what is left. Until then the crew does not pick it again.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs