[processor/resourcedetectionprocessor] Introduce instance life cycle to AWS EC2 resource attributes

[ouyuran]

Description

It's useful to know the EC2 instance life cycle, for example spot, on-demand or reserved. Especially for costing estimation purpose.

As it's not part of semantic conventions, it was added to customized namespace aws.ec2 like what we have for AWS ECS aws.ecs.

We'll be tracking changes to entitiy attributes for k8s in this SemConv issue.

Testing

Has tested the metadata endpoint of EC2.

TOKEN=$(curl -X PUT "http://169.254.169.254/latest/api/token" \
  -H "X-aws-ec2-metadata-token-ttl-seconds: 21600")
curl -H "X-aws-ec2-metadata-token: $TOKEN"   http://169.254.169.254/latest/meta-data/instance-life-cycle

Documentation

Added new attribute to AWS EC2 resource attributes document. It's auto generated.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: ouyuran / name: Yuran Ou (e45f653, d1b66e0, 8468b50, f4ece31)

[ChrsMark]

Is this a state like attribute? I wonder if that should be a metric then 🤔 .

See open-telemetry/semantic-conventions#1554.

Probably discussing this and defining this first with Semantic Conventions project would help.

[ouyuran]

I know the name is confusing, it's actually indicates the EC2 billing model, values can be spot, on-demand or reserved.
I did some research on Semantic Conventions trying to find a place for it but failed. Two possible namespaces

1. cloud, as this attribute is virtual machine specific, but cloud is a generic namespace for any kind of cloud resources, maybe not the best place.
2. host, I guess this one has nothing to do with cloud services, maybe not here.

Therefore, I think maybe we can add a non-standard attribute like what they did for ECS or many other resources, check it out here: https://github.com/open-telemetry/opentelemetry-collector-contrib/blob/main/processor/resourcedetectionprocessor/internal/aws/ecs/documentation.md?plain=1#L11-L16

[ouyuran]

@ChrsMark does it make sense to you, can we move forward this way?

[ChrsMark]

Thank's for the context! I agree that if this attribute cannot be generalised then cloud.* is not appropriate and it should be under aws.* namespace. Could you file an issue in Semantic Conventions for this so it can be moved forward?

@dashpole @Aneurysm9 please share your input as code-owners.

FYI: open-telemetry/opentelemetry-collector#13076

[ouyuran]

Sure, the issue has been created. open-telemetry/semantic-conventions#2300. Apart from aws.ec2.instance_life_cycle, I've also added aws.ec2.vpc_id in the proposal, implementation PR will be created later.

BTW, is e2e required for this change? I've check the test code, since most of them are generated, I'm not sure if I sure do some e2e test, if so where can I find the guideline how to do it?

[ChrsMark]

BTW, is e2e required for this change? I've check the test code, since most of them are generated, I'm not sure if I sure do some e2e test, if so where can I find the guideline how to do it?

I don't think we have e2e tests for this component so probably this addition won't need any.

[ChrsMark]

Sure, the issue has been created. open-telemetry/semantic-conventions#2300. Apart from aws.ec2.instance_life_cycle, I've also added aws.ec2.vpc_id in the proposal, implementation PR will be created later.

Thank's! @dashpole @Aneurysm9 please consider this too as part of this enhancement.

[dashpole]

I'll defer to @Aneurysm9 for changes related to AWS resource detectors

[kaiyan-sheng]

Maybe we can consider instance_life_cycle as an attribute since other cloud providers also have similar pricing models? For example Azure has pay as you go, that would be similar to the on-demand for AWS.