Launching a Burstable Debian 9 ARM Cluster in 1 line with EL2

The burstable instance types for ARM architecture are the T4g variants, as described in the AWS documentation on this subject.


When you ask for a Linux distribution/version on a Graviton instance type like T4g.nano, EL2 has to choose from ARM based AMIs as opposed to the more prevalent Intel x86 based AMIs. If an Intel x86 based AMI was chosen when the requested instance type is T4g.nano, AWS would not make sense of the request, resulting in a failed launch. EL2 does this for all the Graviton instance types: m6g, m6gd, m6gd.metal, c6g, c6gd, c6gd.metal, c6gn, r6g, r6g.metal, r6gd, r6gd.metal, x2gd and x2gd.metal.

So here is how  'Launching a Burstable Debian 9 ARM Cluster in 1 line with EL2'  actually looks like on the command line:

terraform apply -var 'region=ap-south-1' -var 'access_key=...' -var 'secret_key=...' -var 'instance_type=t4g.nano' -var 'instances=3' -var 'linux_distro=debian9stretch' -auto-approve

As you can tell from the value of the region parameter, we're using Asia Pacific South (Mumbai) for this launch, which offers the lowest T4g.nano  On-Demand rates by far.

... and couple more things shown in the above screenshot; in the top right split pane we're ready with a command line to tear the cluster down as soon as needed and in the bottom right split pane we're considering some ways to parameterize our cluster further, such as placing the instances on a custom subnet, or using a partition group strategy, but for this launch that additional functionality isn't used.

So here's how the launch went:

So there you go - a cluster of 3 burstable Graviton ARM based Debian 9  boxes  on the same subnet, up and ready for work. Well, they were for about 10 minutes, but then they were all destroyed as shown in the top right split pane.


REVVOPS EL2 Cluster Pack allows instant clusterization for Layer 7 protocol interoperability by hostname within a cluster, but without the need for a DNS server - enquire now while stocks last.


Cluster placement groups are something different; the cluster placement group strategy, which is one of the three types of placement group strategies, doesn't support burstable instance types, so you couldn't use T4g or even T2 instance types for such cases. Refer to the AWS documentation on 'Cluster placement group rules and limitations' for more details. You can still use Cluster placement group strategies with EL2, but you may only request non burstable instance types.