Serverless monitoring

Your serverless stack, watched from the API to the table

Functions, queues, Step Functions, tables and events on AWS, Azure and GCP, with no code or agents. And the full path of a message, because a 200 from the API does not tell you the work got done.

  • 1 min

    shortest interval for APIs and health endpoints

  • 5 min

    for functions, queues and tables

  • 5 hops

    per end-to-end flow, every 15 min

  • 3 free

    HTTP or heartbeat monitors on any plan

The problem

Why a 200 is not enough in serverless

The API answers, the work does not happen

The API enqueues and returns 200. If the consuming function fails, the ping stays green and nobody notices.

Queues that age silently

Messages pile up with no visible error until a customer asks why nothing arrived.

Crons that stop running

A scheduled task that does not run raises no error: nothing happens, and nobody finds out until much later.

A ping only sees the API

The solution

How AAA Sentinel watches it

  1. Functions, queues and tables, no code

    Lambda, SQS, Step Functions, DynamoDB, EventBridge and API Gateway, and their Azure and GCP equivalents, over the read-only connection: errors, aging messages, throttling and failed deliveries, with adjustable thresholds. From 5 minutes.

  2. Health endpoints read from the inside

    health+json, Spring Actuator, ASP.NET and Kubernetes, or your own JSONPath: if a component goes down, you see it even when the endpoint answers 200. From 1 minute.

  3. The message, end to end

    A flow sends a test transaction into the start of the chain and checks it reaches every hop, up to 5 (e.g. SQS → Lambda → DynamoDB). If it breaks, it tells you which hop and how long each one took. Every 15 minutes.

  4. Crons and confirmed alerts

    A heartbeat waits for your scheduled tasks to check in, from 1 minute to 35 days. When an HTTP probe fails, it is confirmed from Frankfurt when it responds, before you are alerted by email, Slack, Teams, webhook, PagerDuty or Jira Service Management.

At a glance

What it watches on each AWS service

With thresholds you can adjust. Equivalent checks on Azure and GCP, based on what each cloud publishes (on Azure Functions, 5xx responses).

What it watches on each AWS service
ProviderErrorsAging messagesThrottlingFailed deliveries
AWS LambdaYesYesYesNo
Amazon SQSNoYesNoYes
AWS Step FunctionsYesNoNoNo
Amazon DynamoDBNoNoYesNo
Amazon EventBridgeNoNoNoYes
Amazon API GatewayYesNoNoNo

On Lambda, age is that of the stream it processes; on SQS, failed deliveries are messages in its dead-letter queue. On Azure: Functions, Service Bus, Cosmos DB, API Management, Event Grid and Logic Apps. On GCP: Cloud Run, Cloud Functions, Pub/Sub, Eventarc, Workflows, Firestore and API Gateway.

What each plan includes

Free

  • 3 HTTP or heartbeat monitors at no cost
  • Also on Pro and Business, without a pack

Pro · Business

  • Packs of 25 monitors, 3 flows and 1 status page
  • €49/month per pack; €39 from the 4th
  • Health endpoints, cloud resources and flows, with a pack only
See plans

What it does not do

  • It does not check more often than every minute (5 for cloud resources, 15 for flows).
  • It does not run browser tests or Playwright scripts.
  • It does not send SMS or direct calls: it alerts PagerDuty or Jira Service Management.
  • Probes run from Ireland and HTTP failures are confirmed from Frankfurt when it responds; there are no probes in other regions.
  • To test a flow, AAA Sentinel writes a test transaction into your chain.

Frequently asked questions

No. Functions, queues and tables are read through your account connection, and endpoints through their URL. A heartbeat only needs your task to call a URL when it finishes.