Skip to content

MANUAL RUN: copy codeSplitter's data to the codesplitter names and deploy (#809) #825

Closed as not planned

Description

@QuinnsCode

After #809's PR merges, the next deploy has to move the data with it. Until then, any deploy is refused: it would replace codesplitters-db, the three queues and the bucket, and the Alchemy state is still under the old app name.

In your checkout, at a quiet time, with each step done before the next:

cd ~/Documents/projects/RustyBuns && git pull && bun install
cd apps/codesplitter
CLOUDFLARE_ACCOUNT_ID=<codeSplitter account id> bun scripts/move-resources.ts

That one only reads. It should end "Nothing in the way". If it stops, it says why (a Vectorize index that exists, say). In the dashboard, check that the queues codesplitters-hooks, codesplitters-jobs and codesplitters-level-digs have a backlog of 0. Then:

CLOUDFLARE_ACCOUNT_ID=<id> bun scripts/move-resources.ts --copy
CODESPLITTER_DEPLOYS=1 CODESPLITTER_AGENTS=1 CODESPLITTER_CHECKS=1 bunx rustybuns plan

The copy ends with "codesplitter-db matches codesplitters-db". The plan should warn that D1 DB, the queues HOOKS, JOBS and LEVEL_DIGS, and R2 LEVEL_CHUNKS are renamed, and create or delete nothing else. Stop if it wants to create the Worker or a container: that means the state didn't move. Otherwise deploy right away, since writes after the export aren't in the copy:

CODESPLITTER_DEPLOYS=1 CODESPLITTER_AGENTS=1 CODESPLITTER_CHECKS=1 bunx rustybuns deploy --allow-delete

The deploy deletes the old ones. The export stays in apps/codesplitter/.rustybuns/moved/ as a backup. Then sign in on the site and open a repo to check it reads, and update ~/release-codesplitters.sh if it names any of these.

🤖 Generated with Claude Code

Activity

  1. added
    manual runNeeds the owner to run it by hand (tokens, sign-ins); agents can't close it
    on Oct 12, 2026
  2. QuinnsCode commented on Oct 12, 2026

    @QuinnsCode
    OwnerAuthor

    Use zsh ~/release-codesplitters.sh --move for this, not the steps above. Those steps run from ~/Documents/projects/RustyBuns, which has no Alchemy state: the live state is in the agent-office clone's .git/rustybuns/alchemy/apps/codesplitter/state, still under the old stack name codesplitters. So the move script would stop there with "no Alchemy state".

    --move runs from the release worktree and does the steps above in order:

    1. Counts the live site's ready levels (23 on 2026-10-11).
    2. Runs scripts/move-resources.ts read-only.
    3. Opens the Queues page and asks whether the old backlogs are 0.
    4. Asks before running --copy, which copies D1 and moves the state.
    5. Plans. Only DB, HOOKS, JOBS, LEVEL_DIGS and LEVEL_CHUNKS may be replaced; anything else stops it, a recreated Worker or container included.
    6. Deploys with --allow-delete, then checks that the site still lists the same number of ready levels.

    Checked read-only on 2026-10-11: there's no Vectorize index, and none of the codesplitter-* resources exist yet. The plan check was rehearsed against sample plans.

  3. self-assigned this
    on Oct 13, 2026
  4. QuinnsCode commented on Oct 14, 2026

    @QuinnsCode
    OwnerAuthor

    Folded into the one manual-run list, section 9 of #609 (easiest first, with times). This issue's steps stay here; tick the item in #609.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

manual runNeeds the owner to run it by hand (tokens, sign-ins); agents can't close it

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions