Route53¶

Overview¶

• Managed DNS service.
• Most common records -
• A (hostname to IPv4)
• AAAA (hostname to IPv6)
• CNAME (hostname to hostname)
• Alias (hostname to AWS resource)
• Can use public domain names.
• Can use private domain names that can only be resolved by instances within the VPC.
• Costs $0.50/month per hosted zone.
• Global service.
• Provides advanced features like load balancing, health checks, routing policies.

Time to Live (TTL)¶

• Web browser caches DNS record for given number of seconds to minimise DNS lookups.
• Default TTL is 300secs.

Common DNS Records¶

CNAME¶

• Points a hostname to any another hostname.
• Can't be used for root domains.

Alias¶

• Points a hostname to an AWS resource (foo.amazonaws.com)
• Can be used for root domains (zone apex).
• Free of charge.
• Supports native health checks (use the health check configured on the load balancer).
• Can add multiple addresses to the record.

Routing Policies¶

Simple Routing Policy¶

• When you need to redirect to a single resource.
• Doesn't support health checks.
• Can return multiple values to the client, clients choses a random value to use.

Weighted Routing Policy¶

• Controls the percentage of requests that go to particular endpoints.
• Useful way to test new versions of the application on a limited sub-set of requests.
• Can be associated with health checks.

Latency Routing Policy¶

• Direct to the server that has the least latency (based on AWS region).
• A user in Germany might be routed to resources in the US if that has the lowest latency.

Health Checks¶

• By default, health state changes after 3 successful/unsuccessful health checks in a row.
• 30second health check interval by default. Fast health checks happen every 10seconds but costs money.
• Around 15 health checkers will check the endpoint health. So roughly 1 request/sec.
• Supports HTTP, TCP, HTTPS (no SSL verification) healthchecks.
• Can integrate healthcheck with CloudWatch.
• Can use string matching, invert the health check status, and select which regions the healthcheck should run from.
• Can monitor endpoint, cloudwatch alarm, or do a calculated healthcheck.

Failover Routing Policy¶

• Failover to secondary site when healthcheck fails.
• Can only have one primary, and one secondary record (duh).

GeoProximity Routing Policy¶

• Routed based on user location.
• Need to have a default policy for users in a location that doesn't have an explicit policy.
• Increase/decrease bias value to control how much traffic is sent to each region based on where the user is.

Multi-value Routing Policy¶

• Route traffic to multiple resources
• Each multi-value record is associated with a route53 health check.
• Up to 8 healthy records returned per multi-value query.
• If one resource becomes unavailable, it will be removed from the returned DNS record, and the client can chose one of the remaining available resources.