fix(cron): retry and serialize cleanup dispatch on the task definition

Author: TheodoreSpeaks

apps/sim/app/api/cron/cleanup-file-versions/route.ts

@@ -3,6 +3,7 @@ import { type NextRequest, NextResponse } from 'next/server'
 import { verifyCronAuth } from '@/lib/auth/internal'
 import { getJobQueue } from '@/lib/core/async-jobs'
 import { withRouteHandler } from '@/lib/core/utils/with-route-handler'
+import { CLEANUP_DISPATCH_MAX_ATTEMPTS } from '@/background/cleanup-dispatch'
 
 export const dynamic = 'force-dynamic'
 
@@ -21,7 +22,7 @@ export const GET = withRouteHandler(async (request: NextRequest) => {
       'cleanup-dispatch',
       { jobType: 'cleanup-file-versions' },
       {
-        maxAttempts: 1,
+        maxAttempts: CLEANUP_DISPATCH_MAX_ATTEMPTS,
         jobId: `cleanup-dispatch:cleanup-file-versions:${scheduleWindow}`,
         name: 'File version cleanup dispatch',
         concurrencyKey: 'cleanup-dispatch:cleanup-file-versions',

apps/sim/background/cleanup-dispatch.ts

@@ -5,12 +5,17 @@ export interface CleanupDispatchPayload {
   jobType: CleanupJobType
 }
 
+/** Attempts per dispatch. A rerun only re-triggers chunks that delete already-expired data. */
+export const CLEANUP_DISPATCH_MAX_ATTEMPTS = 3
+
 /**
  * Resolves a retention job's workspace scope and fans out its per-chunk cleanup runs. The scan
- * covers every active workspace, so it runs here rather than inside the cron request.
+ * covers every active workspace, so it runs here rather than inside the cron request. One
+ * dispatch at a time per job type: the route keys runs by `concurrencyKey`.
  */
 export const cleanupDispatchTask = task({
   id: 'cleanup-dispatch',
-  retry: { maxAttempts: 1 },
+  queue: { concurrencyLimit: 1 },
+  retry: { maxAttempts: CLEANUP_DISPATCH_MAX_ATTEMPTS },
   run: ({ jobType }: CleanupDispatchPayload) => dispatchCleanupJobs(jobType),
 })