Catalyst Job Scheduling

Job Scheduling runs background work in three pieces: job pools (capacity containers), jobs (one execution each, submitted immediately via API/SDK/MCP), and crons (schedulers that submit jobs on a timetable). Every fact in this skill and its references is runtime-verified against a live Catalyst project (Aug 2026) unless marked otherwise.

Prerequisites

Before any operation, load references/job-scheduling-basics.md — it covers the five setup rules that cause most failures: job pool must exist first (no default), targets must be JOB-type functions, all time values in seconds, job_name limits, and function-vs-pool memory constraints.

How It Works

1. Identify the target type: Job function (most common), Webhook (any HTTP URL), Circuit, or AppSail. Create/verify a matching-type job pool (Function pools size by memory MB; the others by concurrent number 1–10).
2. One-off work → submit an immediate job (MCP CatalystbyZoho_Create_Immediate_Job or SDK jobScheduling().job().submitJob()). Recurring or delayed work → create a cron (CatalystbyZoho_Create_Cron_Job or SDK cron().createCron()) choosing Periodic / OneTime / Calendar / CronExpression. Load references/job-scheduling-basics.md for exact payloads for both paths.
3. Write the Job function handler (jobRequest, context): initialize the SDK with catalyst.initialize(context, { scope: 'admin' }), read inputs via jobRequest.getAllJobParams() (all values arrive as strings), and ALWAYS end with context.closeWithSuccess() or closeWithFailure() — failure is what triggers the retry chain.
4. For retries, dynamic scheduling, webhook targets, or monitoring, load references/job-scheduling-advanced.md — retry semantics (new job records linked by parent_job_id), firing behaviors (Periodic crons fire immediately on create AND update; past-dated OneTime fires instantly; OneTime auto-disables after firing), and observability traps (success/failure counters stay 0; dynamic crons are hidden from list APIs).
5. Verify by job records, not counters: CatalystbyZoho_Get_Job_By_Id shows job_status (PENDING → RUNNING → SUCCESS | FAILURE), execution_time, dispatch_delay, response_code. Do not trust cron success_count/failure_count — they stay 0.

Security Checklist