fix(db): repair partial member sync status on existing databases

Author: Sg312

.github/workflows/test-build.yml

@@ -210,7 +210,7 @@ jobs:
         working-directory: packages/db
         env:
           KNOWLEDGE_ACL_TEST_DATABASE_URL: postgresql://postgres:postgres@127.0.0.1:5432/sim_auth_scim
-        run: bunx vitest run script-migrations/0016_backfill_search_vectors.postgres.test.ts
+        run: bunx vitest run script-migrations/0016_backfill_search_vectors.postgres.test.ts member-sync-status-migration.postgres.test.ts
 
       - name: Verify Search progress, pagination, and outbox scheduling in PostgreSQL
         working-directory: apps/sim

packages/db/member-sync-status-migration.postgres.test.ts

@@ -0,0 +1,79 @@
+import { readFile } from 'node:fs/promises'
+import { generateId } from '@sim/utils/id'
+import postgres, { type Sql } from 'postgres'
+import { afterAll, beforeAll, describe, expect, it } from 'vitest'
+
+const databaseUrl = process.env.KNOWLEDGE_ACL_TEST_DATABASE_URL
+
+describe.runIf(Boolean(databaseUrl))('member sync status upgrade in PostgreSQL', () => {
+  let admin: Sql
+  let sql: Sql
+  let migration: string
+  const schema = `member_sync_status_${generateId().replaceAll('-', '')}`
+
+  beforeAll(async () => {
+    const url = new URL(databaseUrl!)
+    if (
+      !['localhost', '127.0.0.1'].includes(url.hostname) ||
+      !/^\/sim_(acl_test|auth_scim)(_|$)/.test(url.pathname)
+    ) {
+      throw new Error('Member sync migration tests require a disposable local test database')
+    }
+    admin = postgres(url.toString(), { max: 1, onnotice: () => undefined })
+    await admin.unsafe(`CREATE SCHEMA "${schema}"`)
+    sql = postgres(url.toString(), {
+      max: 1,
+      connection: { search_path: schema },
+      onnotice: () => undefined,
+    })
+    const original = await readFile(
+      new URL('./migrations/0319_permission_aware_knowledge.sql', import.meta.url),
+      'utf8'
+    )
+    const table = original.match(
+      /CREATE TABLE IF NOT EXISTS "knowledge_connector_member_sync_log" \([\s\S]*?\n\);/
+    )?.[0]
+    if (!table) throw new Error('Original member sync log table DDL was not found')
+    await sql.unsafe(table)
+    migration = await readFile(
+      new URL('./migrations/0356_member_sync_partial_status.sql', import.meta.url),
+      'utf8'
+    )
+  })
+
+  afterAll(async () => {
+    await sql?.end()
+    if (admin) {
+      await admin.unsafe(`DROP SCHEMA IF EXISTS "${schema}" CASCADE`)
+      await admin.end()
+    }
+  })
+
+  it('repairs partial completion after an account failure without losing run history', async () => {
+    await sql`INSERT INTO knowledge_connector_member_sync_log
+      (id, connector_id, status, members_claimed, members_failed)
+      VALUES ('current', 'connector', 'started', 1, 1),
+             ('history', 'connector', 'completed', 1, 0)`
+    await expect(
+      sql`UPDATE knowledge_connector_member_sync_log SET status = 'partial' WHERE id = 'current'`
+    ).rejects.toMatchObject({ code: '23514', constraint_name: 'kcmsl_status_check' })
+
+    await sql.unsafe(migration)
+    await sql`UPDATE knowledge_connector_member_sync_log SET status = 'partial' WHERE id = 'current'`
+    await sql.unsafe(migration)
+    expect(
+      await sql`SELECT id, status, members_failed FROM knowledge_connector_member_sync_log ORDER BY id`
+    ).toEqual([
+      { id: 'current', status: 'partial', members_failed: 1 },
+      { id: 'history', status: 'completed', members_failed: 0 },
+    ])
+    for (const status of ['started', 'completed', 'failed']) {
+      await sql`INSERT INTO knowledge_connector_member_sync_log (id, connector_id, status)
+        VALUES (${status}, 'connector', ${status})`
+    }
+    await expect(
+      sql`INSERT INTO knowledge_connector_member_sync_log (id, connector_id, status)
+        VALUES ('invalid', 'connector', 'unknown')`
+    ).rejects.toMatchObject({ code: '23514', constraint_name: 'kcmsl_status_check' })
+  })
+})

packages/db/migrations/0356_member_sync_partial_status.sql

@@ -0,0 +1,7 @@
+-- Older deployments already journaled 0326 before it included the partial status.
+-- Widen the existing check atomically; every previously valid status remains valid.
+-- migration-safe: Atomic widening of the 0319 status check to match the already-deployed partial writer from 0326; all old statuses remain accepted and no rows or columns are removed.
+ALTER TABLE "knowledge_connector_member_sync_log"
+  DROP CONSTRAINT IF EXISTS "kcmsl_status_check",
+  ADD CONSTRAINT "kcmsl_status_check"
+    CHECK ("status" IN ('started', 'partial', 'completed', 'failed')) NOT VALID;