AWS Cloud Map service health check configuration

Health checks help determine whether service instances are healthy or not. If you don't configure a health check during service creation, traffic will be routed to service instances regardless of the instances' health status. When you configure a health check, AWS Cloud Map returns healthy resources by default. You can use the HealthStatus parameter of the DiscoverInstances API to filter resources by health status and get a list of unhealthy resources. You can also use the GetInstancesHealthStatus API to retrieve the health status of a particular service instance.

You can either configure a Route 53 health check or a custom, third-party health check when you create an AWS Cloud Map service.

Route 53 health checks

If you specify settings for an Amazon Route 53 health check, AWS Cloud Map creates a Route 53 health check whenever you register an instance and deletes the health check when you deregister the instance.

For public DNS namespaces, AWS Cloud Map associates the health check with the Route 53 record that AWS Cloud Map creates when you register an instance.If you specify both A and AAAA record types in a service's DNS configuration, AWS Cloud Map creates a health check that uses the IPv4 address to check the health of the resource. If the endpoint that's specified by the IPv4 address is unhealthy, Route 53 considers both the A and AAAA records to be unhealthy. If you specify a CNAME record type in a service's DNS configuration, you can't configure a Route 53 health check.

For namespaces that you use API calls to discover instances for, AWS Cloud Map creates a Route 53 health check. However, there's no DNS record for AWS Cloud Map to associate the health check with. To determine whether a health check is healthy, you can configure monitoring using either the Route 53 console or using Amazon CloudWatch. For more information about using the Route 53 console, see Get Notified When a Health Check Fails in the Amazon Route 53 Developer Guide. For more information about using CloudWatch, see PutMetricAlarm in the Amazon CloudWatch API Reference.

Note

• You can't configure an Amazon Route 53 health check for a service created in a private DNS namespace.

• A Route 53 health checker in each health-checking AWS Region sends a health check request to an endpoint every 30 seconds. On average, your endpoint receives a health check request about every two seconds. However, health checkers don't coordinate with one another. Therefore, you might sometimes see several requests in one second that's followed by a few seconds with no health checks at all. For a list of health-checking regions, see Regions.

For information about the charges for Route 53 health checks, see Route 53 Pricing.

Custom health checks

If you configure AWS Cloud Map to use a custom health check when you register an instance, you must use a third-party health checker to evaluate the health of your resources. Custom health checks are useful in the following circumstances:

• You can't use a Route 53 health check because the resource isn't available over the internet. For example, suppose that you have an instance that's located in an Amazon VPC. You can use a custom health check for this instance. However, for the health check to work,your health checker must also be in the same VPC as your instance.

• You want to use a third-party health checker regardless of where your resources are.

When you use a custom health checks, AWS Cloud Map doesn't check the health of a given resource directly. Instead, the third-party health checker checks the health of the resource and returns a status to your application. Your application will then need to submit a UpdateInstanceCustomHealthStatus request that relays this status to AWS Cloud Map. If the initial status relayed is UNHEALTHY, and if there isn't another UpdateInstanceCustomHealthStatus within 30 seconds that relays a status of HEALTHY, the resource is confirmed to be unhealthy. AWS Cloud Map stops routing traffic to that resource.