Enriching AWS chaos

[STRRL]

Here is an overview of the plan for supporting more chaos about AWS.

We have already try to implement AWS Chaos like "stop/restart" EC2, but it still lacks e2e tests. Thanks to localstack, we could emulation the AWS API at the local machine. So we could write some test cases based on localstack.

Except for e2e tests, I think we could implement more chaos about AWS based services, such as S3 or DynamoDB.

[fewdan]

I think this is a very good idea!

[someshkoli]

Hey @STRRL, I'd like to work on this issue

Here is an overview of the plan for supporting more chaos about AWS.

Except for e2e tests, I think we could implement more chaos about AWS based services, such as S3 or DynamoDB.

From what I understand, We need to use the api that awscli uses to create more chaos on aws services such just like we do on EC2 ?

We have already try to implement AWS Chaos like "stop/restart" EC2, but it still lacks e2e tests. Thanks to localstack, we could emulation the AWS API at the local machine. So we could write some test cases based on localstack.

Not yet clear about this, could you please elaborate more ?

[STRRL]

We need to use the api that awscli uses to create more chaos on aws services such just like we do on EC2?

It's the way that we used to implement AWS Chaos for now. We invoke the official AWS API for making some chaos. It indeed makes some breaks on the real machines.

If we could, we want to try another way to make more chaos: use localstack instead of official AWS API. You might already notice that awscli could connect to another self-hosted localhost by providing --endpoint.

We might run a modified localstasck that could provide some internal API for simulation chaos. But it's just an imagination for injecting chaos into "self-hosted cloud infrastructure".

Honstly, if we modify the localstack, these chaoses might not work on AWS, it might be useless. We will figure out the feasibility and usability in several days.

[STRRL]

And this issue has been taken as an lfx mentorship project, you could look up here. 😜

[someshkoli]

We need to use the api that awscli uses to create more chaos on aws services such just like we do on EC2?

It's the way that we used to implement AWS Chaos for now. We invoke the official AWS API for making some chaos. It indeed makes some breaks on the real machines.

If we could, we want to try another way to make more chaos: use localstack instead of official AWS API. You might already notice that awscli could connect to another self-hosted localhost by providing --endpoint.

We might run a modified localstasck that could provide some internal API for simulation chaos. But it's just an imagination for injecting chaos into "self-hosted cloud infrastructure".

Honstly, if we modify the localstack, these chaoses might not work on AWS, it might be useless. We will figure out the feasibility and usability in several days.

Okay got it, do we already have a POC for this ? If not I shall start by making one.

[someshkoli]

And this issue has been taken as an lfx mentorship project, you could look up here. 😜

Great, I'll apply for this ✌️

[STRRL]

Okay got it, do we already have a POC for this? If not I shall start by making one.

No, we have not built out any POC. Thanks a lot! 😍

[supra08]

Hi @STRRL!

I am new to the chaos-mesh community and also willing to apply to the LFX mentorship.

I had a couple of questions regarding this task. You mentioned:

We might run a modified localstack that could provide some internal API for simulation chaos.

1. Why do we need a 'modified' localstack? Because localstack seems to provide AWS cloud mocking tools out of the box with the API endpoint given in https://github.com/localstack/localstack#configurations. Do we want a chaos simulation module over this given API?

2. If this is to be a separate module, would it be a separate service in the chaos infrastructure?

3. I see that a PR is present for the integration test: awschaos: add integration test with localstack #1477. Does this PR validate the localstack integration and we can proceed the module POC?

[STRRL]

Why do we need a 'modified' localstack?

Modifying localstack is one way for injecting chaos, it doesn't mean we must do it.

If this is to be a separate module, would it be a separate service in the chaos infrastructure?

I'm not sure about what the "service" means. It mostly like DNSChaos, it contains the standalone Kubernetes Service stands for localstack, and separate CRD.

I see that a PR is present for the integration test: #1477. Does this PR validate the localstack integration and we can proceed the module POC?

Well, it's one of the test cases. But it does not bring new chaos. It shows the way that localstack could integrate easily with our test framework.

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 21 days

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 21 days