Monitoring scheduled AWS Lambda functions (EventBridge)

Scheduled Lambdas (EventBridge Scheduler or an EventBridge rule) fail quietly in ways a metric alarm can miss. The schedule gets disabled or deleted in a deploy, the function times out before finishing, a permissions change stops the invocation, or the code catches an exception and returns normally. A heartbeat checks the one thing you care about: did the job finish on time?

Python runtime

import os
import urllib.request

PING = os.environ["PING_URL"]


def ping(suffix=""):
    try:
        urllib.request.urlopen(PING + suffix, timeout=5).close()
    except Exception as exc:  # never let monitoring break the job
        print(f"heartbeat ping failed: {exc}")


def handler(event, context):
    ping("/start")
    try:
        do_the_work()
    except Exception:
        ping("/fail")
        raise
    ping()

Node.js runtime (18+)

const ping = (s = "") =>
  fetch(process.env.PING_URL + s, { signal: AbortSignal.timeout(5000) }).catch(() => {});

export const handler = async () => {
  await ping("/start");
  try {
    await doTheWork();
  } catch (err) {
    await ping("/fail");
    throw err;
  }
  await ping();
};

Match the schedule

EventBridge writes cron with six fields, for example cron(0 3 * * ? *). In CronPup, use the standard five-field form, 0 3 * * *. EventBridge rules run in UTC, so set the monitor's timezone to UTC. EventBridge Scheduler lets you pick a timezone, so use the same one in CronPup.

Timeouts and retries

• If the function hits its timeout, it never sends the success ping, so the monitor goes down once the grace period passes. Set the grace period a little above the function timeout.
• Asynchronous invocations are retried by Lambda (up to two retries by default). A retry that succeeds pings success and marks the monitor up again.
• The function needs outbound internet access. Inside a VPC, that means a NAT gateway.

Monitor your scheduled Lambdas. 10 monitors are free.