Migrate e2e test for AWS SQS Queue scaler to Go.

[v-shenoy]

Proposal

Relates to #2357, #2737.

Use-Case

No response

Anything else?

No response

[aviadlevy]

Hi,
I might be interested in working on this issue, but first needs to verify I understand the current behvior

In e2e test you querying an actual SQS queue in AWS? not mock?

[v-shenoy]

Hi, I might be interested in working on this issue, but first needs to verify I understand the current behvior

In e2e test you querying an actual SQS queue in AWS? not mock?

Yes, all of the e2e tests query actual resources.

[aviadlevy]

So I'll start take a look at this test. you can assign this to me 😎

[v-shenoy]

Ahaha, I am not in charge here :) Just helping out. Think you can assign it to yourself as well actually.

[aviadlevy]

Can't assign myself, so let's wait for KEDA owners

[aviadlevy]

I'm still wondering about using real AWS resources. don't you think it's better to use sqs mock? (something like https://hub.docker.com/r/roribio16/alpine-sqs/)

My concern is case when something really bad happening and we aren't able to clean the AWS resource created by the test. how do you handle such cases?

Of course a pre condition requirement is the ability to set custom endpoint for aws resources which is currently unsupported.

Let me know what you think.
@v-shenoy @JorTurFer

[JorTurFer]

hey @aviadlevy
in fact, we prefer to use real resources for e2e always, where is possible. If e2e test fail and the resources are not deleted, we can delete them manually, going to the cloud provider console to do it 😄
if the test fails without a panic, the resources should be deleted, and even that, after solving the panic, the resources will be deleted (running and failing the test again).

The advantage of using real resources is that we can detect/check real behaviors, mocks usually doesn't support all the scenarios like authentication or pod-identities (we want to cover them also with e2e test).

From costs pov, CNCF gives credit on principal cloud providers so it's not a problem, we try to cleanup all resources after tests but it's not a problem if they are there after the test during some time due to an error

[aviadlevy]

Thanks @JorTurFer for the elaborate explanation