Worker orchestration

How the worker schedules monitors, handles retries, and coordinates locks.

The worker pulls jobs from Redis-backed queues, enforces rate limits, and writes snapshots atomically.

Scheduling

• Each monitor stores nextRunAt; jobs are enqueued accordingly.
• Locks prevent double-processing when multiple workers run.
• Grace windows adjust cadence automatically based on plan state.

Retries

• Transient errors are retried with exponential backoff and jitter.
• Retry-After headers on 429 are honored before continuing.

for (let attempt = 0; attempt <= retryMaxAttempts; attempt++) {
  last = await attemptOnce();
  if (!last.error && last.statusCode && last.statusCode < 400) break;
  const backoff = retryBaseDelayMs * Math.pow(2, attempt);
  await sleep(Math.min(backoff + Math.random() * retryBaseDelayMs, 30_000));
}

Processing pipeline

• Fetch target → normalize → hash → compare → write snapshot/diff → send alerts/webhooks → update health.
• Health evaluation uses recent snapshots to avoid flapping.