Regions and zones

So QuickCart needs to know two things: where AWS’s computers physically are, and how to spread across them so that one building failing is a non-event.

AWS isn’t one giant computer in one place. It’s many buildings, in many parts of the world, grouped in two levels: regions and availability zones.

A region is a part of the world

A region is a separate area where AWS runs a complete copy of its services, such as Northern Virginia, Ireland or Tokyo. Each has a code:

There are more than 30 regions, and AWS adds more every year. Regions are kept deliberately separate from each other, so a problem in one rarely spreads to another. Most things you create belong to one region and are invisible from the others.

You choose a region for four reasons:

This course uses us-east-1 everywhere. It’s one of the cheapest regions, and it has every service.

An availability zone is a building, or a few

Inside each region are several availability zones, usually called zones or AZs. A zone is one or more data centres with their own power, cooling and network. AWS keeps zones in the same region miles apart, so a fire, flood or power cut hits only one. They’re also close enough (within about 100 km of each other) to be joined by fast private links. A server in one zone can talk to a server in another in a millisecond or two.

us-east-1 has six zones, named us-east-1a to us-east-1f.

What lives where

Every AWS resource is one of three kinds. You need to know which kind a resource is before you can spread it safely:

The zonal ones are where people get caught. A server is zonal, so two servers in one zone share its fate. Spreading them is your job.

Spreading QuickCart across two zones

For QuickCart, “two zones” means a matching set of everything zonal in each:

When zone a fails, the load balancer notices that the servers there stopped answering its health checks and sends everything to zone b. The load balancers lesson shows how. The database switches to its standby, which the RDS lesson covers.

Two things make this work, and both cost something:

Zone names and zone IDs

Here’s a detail that confuses people. In us-east-1 and other older regions, AWS shuffles the zone names for each account. Your us-east-1a might be a different building from someone else’s us-east-1a. AWS did this so that everybody picking “a” didn’t pile into the same building.

Each zone also has a zone ID, like use1-az2, that means the same building for everyone:

You’ll mostly use names. Reach for IDs whenever two AWS accounts need to agree on a zone.

Zones versus regions

Two zones protects you against one building failing. That’s by far the most common kind of trouble. It doesn’t protect against a problem across a whole region, like the 2017 S3 outage. For that you’d run a second copy of QuickCart in another region. That’s much more work, so most teams do it only when the business needs it. Environments and regions shows how Terraform makes it practical.

Below, on the left, are some of AWS’s regions, with us-east-1 circled. On the right is what’s inside us-east-1: three of its zones, QuickCart’s two servers, and the load balancer in front. Send some orders. Then fail us-east-1a and send them again. Then switch to Two zones and repeat. Try failing the other zones too, and two zones at once.

What to notice: in both setups QuickCart had two servers. The only difference was where they ran. Failing a zone that has nothing of yours in it changes nothing. Failing the zone that has everything takes QuickCart down.

Free — you only look around. Nothing in this lab creates anything.

You’ll find the region selector, list the zones in us-east-1 with their zone IDs, and see which things are global, regional and zonal. About 10 minutes. You need to be signed in to the AWS console.

1. Find the region selector

   Look at the top-right corner of the console, next to your account name. The region selector shows the name of the region you’re working in, such as N. Virginia.

   Open it.

   You should see

   A list of regions, each with its name and code, such as US East (N. Virginia) us-east-1 and Europe (Ireland) eu-west-1. Some near the bottom may be marked as not enabled. Newer regions are switched off until you opt in.

   Choose US East (N. Virginia) us-east-1. Everything in this course happens there.

2. List the zones and their IDs

   AWSIn the AWS console: EC2→Dashboard

   Find the Zones table. It’s in the dashboard’s service health section.

   You should see

   Six rows, us-east-1a to us-east-1f, each with a Zone ID from use1-az1 to use1-az6. Which name goes with which ID is specific to your account. Write down which ID your us-east-1a has. A colleague’s may be different.

   If the console layout has moved, look for any table with Zone name and Zone ID columns. In the EC2 console, Settings > Zones shows the same list.

3. See one subnet per zone

   Every account comes with a default VPC in each region, and it has one subnet in each zone. That makes it an easy way to see zones in action.

   AWSIn the AWS console: VPC→Subnets

   You should see

   Six subnets belonging to the default VPC, one per zone. The Availability Zone column shows us-east-1a to us-east-1f.

   If there’s no Availability Zone ID column, choose the gear icon above the table, switch Availability Zone ID on, and choose Confirm. Each subnet now shows its zone ID as well. It matches the table from the last step.

4. See that regions are separate

   Open the region selector and choose Europe (Ireland) eu-west-1. Stay on VPC > Subnets.

   You should see

   A different set of default subnets, in zones named eu-west-1a, eu-west-1b and eu-west-1c, with IDs starting euw1-. Nothing from us-east-1 appears here.

   Switch the region back to US East (N. Virginia) us-east-1.

5. See what's global

   AWSIn the AWS console: IAM→Dashboard

   Look at the region selector.

   You should see

   It says Global and you can’t change it. Users and roles aren’t in any region. A role works the same in us-east-1, eu-west-1 and everywhere else.

6. Find where AWS reports problems

   Type Health into the search bar at the top of the console and open AWS Health Dashboard. Choose Service health. The wording of this menu changes from time to time.

   You should see

   A list of open and recent issues by service and region. When something odd is happening in one zone or region, AWS reports it here. Check it early during an incident, before assuming your own code is to blame.

7. Clean-up step: Clean up

   This lab didn’t create anything, so there’s nothing to delete.

   Check that the region selector shows N. Virginia (us-east-1) again, so the next lesson starts in the right place.

   You should see

   The region selector reads N. Virginia.