supabase/apps/studio/data/ha-admin/keys.ts at create-pull-request/patch - supabase - Gitea: 个人代码仓库

lywsvip/supabase

mirror of https://github.com/supabase/supabase.git synced 2026-09-07 02:20:52 +08:00

Files

claude[bot] 10950d286b chore(studio): address review comments on Multigres topology diagram (#49592 )

_Requested by **Alaister Young** · [Slack
thread](https://supabase.slack.com/archives/C0161K73J1J/p1787738517181409?thread_ts=1787635785.354489&cid=C0161K73J1J)_

Follow-up to #49298, which was squash-merged before @joshenlim's last
review round was addressed. Picking up the review comments here.

## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.

YES

## What kind of change does this PR introduce?

Chore — dead code removal and comment corrections. No behavior change.

## What is the current behavior?

Three of @joshenlim's review comments on #49298 are still open on
master:

- [Dead code in
`ha-cluster-cells-query.ts`](https://github.com/supabase/supabase/pull/49298#discussion_r3861029414)
— "seems to be dead code? no one's consuming this file"
- [Dead code in
`ha-cluster-databases-query.ts`](https://github.com/supabase/supabase/pull/49298#discussion_r3861031230)
— "likewise - seems to be dead code"
- [`STATUS_BADGE_VARIANTS`
statuses](https://github.com/supabase/supabase/pull/49298#discussion_r3861095281)
— "just to sanity check these are the only statuses? are there any
failure states? e.g 'Failed'"

Concretely, on master today:

- `apps/studio/data/ha-admin/ha-cluster-cells-query.ts` and
`apps/studio/data/ha-admin/ha-cluster-databases-query.ts` ship query
options that nothing imports. The diagram only reads `/poolers` and
`/gateways`.
- `multipoolerSchema.lifecycleStatus` is an undocumented `z.string()`,
while its neighbours `type` and `servingStatus` both list their expected
proto values in a comment.
- The comment above `HA_POOLER_STATUS_LABELS` claims the labels "Matches
the status vocabulary of the read replica surfaces (getStatusLabel)".
They don't — `getStatusLabel` in `ReadReplicas/ReadReplicas.utils.ts`
also returns `Failed`, `Restarting`, `Resizing` and `Restoring`, none of
which the HA labels have.

## What is the new behavior?

- Deleted both dead query modules and pruned the orphaned `cells` and
`databases` factories from `haAdminKeys`, keeping `poolers` and
`gateways`. Verified by grep that neither file name nor any of their
exported symbols (`haClusterCellsQueryOptions`, `HaClusterCellsData`,
`haClusterDatabasesQueryOptions`, `HaClusterDatabasesData`, the
`*Variables`/`*Error` types) nor `haAdminKeys.cells` /
`haAdminKeys.databases` has a single reference left anywhere outside the
deleted files. No re-export shims left behind. `get-ha-admin.ts` stays —
`poolers` and `gateways` still use it.
- Documented `lifecycleStatus` against the actual enum,
`PoolerLifecycleStatus` in [multigres
`proto/clustermetadata.proto`](https://github.com/multigres/multigres/blob/main/proto/clustermetadata.proto):
`LIFECYCLE_UNKNOWN` (zero value, omitted from JSON) | `STARTING` |
`ACTIVE` | `STOPPING` | `SHUTDOWN` | `QUARANTINED`. `getPoolerStatus`
already maps every member.
- Reworded the `HA_POOLER_STATUS_LABELS` comment to say the labels are a
subset drawn from the read replica vocabulary rather than a match for
it, and noted where the read replica `Failed` lands on the HA side.

**On the `Failed` question:** the answer from the proto is that there is
no dedicated failure member. The terminal states are `QUARANTINED` — the
pooler "has given up trying to become a healthy replica: it cannot
automatically recover to a functioning state (e.g. it could not complete
a pg_rewind, could not restore from backup to start postgres, or fell
irrecoverably behind on replication)", kept alive for forensics — and
`SHUTDOWN`, "durably down". Both already map to `unhealthy` / the
`Unhealthy` warning badge, so the four statuses on
`STATUS_BADGE_VARIANTS` are complete for the enum as it stands. If we'd
rather show `QUARANTINED` as its own `Failed` status with a destructive
badge (matching the read replica surface), that's a small follow-up — a
product/copy call rather than a gap, so not folded in here.

**Not included: [the replication page UX
comment](https://github.com/supabase/supabase/pull/49298#discussion_r3861055216)**
("is there any other content we plan to add here? it feels empty atm...
it's just a repeat of the home page + settings/infrastructure").
@joshenlim flagged that one himself as "UX feedback which can be
addressed separately". It's a product and IA question about what that
page is for, not something to answer with a code change here — leaving
it for @alaister and design.

## Additional context

- Verified locally: `tsc --noEmit` (0 errors), ESLint on the touched
files (clean), Prettier check (clean), and `HaTopology.utils.test.ts` +
`HaInstanceConfiguration.utils.test.ts` (26/26 passing). CI is green as
well.
- Exhaustive grep across the repo (excluding `node_modules`/`.git`/build
output, covering `apps/**` incl. `lite-studio`, `packages/**` and
`e2e/**`) confirmed zero remaining references to the deleted files,
their exported symbols, and the removed key factories.
- No test changes: the diff deletes unreferenced code and edits comments
only, so there's no new behavior to cover. `HaTopology.utils.test.ts`
already pins every `lifecycleStatus` value listed in the new comment.

---
_Generated by [Claude
Code](https://claude.ai/code/session_012StGVQSmPzpGTXrduo9Xyi)_

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Alaister Young <10985857+alaister@users.noreply.github.com>

2026-08-27 08:52:52 +00:00

7 lines

251 B

TypeScript

Raw Permalink Blame History

 export const haAdminKeys = {
   poolers: (projectRef: string | undefined) =>
     ['projects', projectRef, 'ha-admin', 'poolers'] as const,
   gateways: (projectRef: string | undefined) =>
     ['projects', projectRef, 'ha-admin', 'gateways'] as const,
 }