cron
@rtorcato/cf-common/cron builds the scheduled handler for a Worker with Cron
Triggers. You map each cron expression to a job, and the handler runs the one
that fired.
import { defineScheduled } from '@rtorcato/cf-common/cron'
export default {
scheduled: defineScheduled<Env>({
'0 * * * *': async (_controller, env) => rollupHourly(env),
'0 0 * * *': async (_controller, env) => rollupDaily(env),
}),
}
Configure the triggers
The keys must match the crons in wrangler.jsonc exactly, character for
character:
"triggers": {
"crons": ["0 * * * *", "0 0 * * *"]
}
The job
Each job gets the same arguments as a scheduled handler:
'*/5 * * * *': async (controller, env, ctx) => {
console.log(controller.cron, new Date(controller.scheduledTime).toISOString())
ctx.waitUntil(flushMetrics(env))
}
controller.cron is the expression that fired and controller.scheduledTime is
when it was due, in milliseconds.
Alongside fetch
defineFetch returns an object, so spread it to put both handlers in one Worker:
const api = defineFetch<Env>(handler)
export default {
...api,
scheduled: defineScheduled<Env>({ '0 * * * *': hourly }),
}
When nothing matches
A trigger with no job throws a CloudflareError with code CRON_NO_MATCH. It
shows up in the Worker's logs, which is where a typo between the config and the
code gets caught. An error thrown by a job propagates the same way.
Test it locally
wrangler dev --test-scheduled exposes a route that fires the handler:
curl "http://localhost:8787/__scheduled?cron=0+*+*+*+*"
For a runnable walkthrough, see the scheduled handler in apps/example. See the
cron API reference for every option.