Add 'platformName' to selenium-grid scaler metadata structure

[gifflet]

Proposal

Currently, the selenium grid scaler only supports browserName and browserVersion capabilities in trigger metadata. This results in the impossibility of configuring a trigger for tests targeting specific platforms like Windows 10, Windows 11, etc.

Use-Case

If the platformName capability were supported it would be possible to configure a trigger like this:

triggers:
  - type: selenium-grid
    metadata:
      url: 'http://selenium-hub:4444/graphql'
      browserName: 'chrome'
      browserVersion: '108.0'
      unsafeSsl: 'true',
      platformName: 'Windows 10'

Anything else?

The possibility of defining the platformName would make sense if it were possible to run a virtual machine, equivalent to the platform defined in the capability platformName, in the pod. In Linux this is possible through the native feature called KVM (Kernel-based Virtual Machine). That is, the operating strategy would be to make this resource available to the linux container (through /dev/kvm) and within the container use qemu, for example, to run the virtual machine.

[JorTurFer]

Hi,
Basically you would like to have the plaformName field and filter by it in the selenium scaler, right? I mean, currently KEDA checks the queue size for the given browserName and you would like to use both parameters, browserName and platformName to be more accurate with the queue length for the specific case. Am I correct?

I don't understand the part of support to assign /dev/kvm linux device to spawned pods. KEDA doesn't assign anything, I mean, you have your workload and KEDA just expose metrics, if you want to assign/mount/whatever that path in your selenium agents, you have to do in your manifest, KEDA doesn't patch them anyhow. I feel that I'm missing something

[gifflet]

Basically you would like to have the plaformName field and filter by it in the selenium scaler, right? I mean, currently KEDA checks the queue size for the given browserName and you would like to use both parameters, browserName and platformName to be more accurate with the queue length for the specific case. Am I correct?

Yeah, It's right.

I don't understand the part of support to assign /dev/kvm linux device to spawned pods. KEDA doesn't assign anything, I mean, you have your workload and KEDA just expose metrics, if you want to assign/mount/whatever that path in your selenium agents, you have to do in your manifest, KEDA doesn't patch them anyhow. I feel that I'm missing something

As it is necessary to associate the /dev/kvm device to the pod so that it can be possible to run a virtual machine inside the container, what procedure would you recommend for such a configuration?

[JorTurFer]

No no, I'm not saying that the way isn't that, I'm just saying that KEDA doesn't manage it. KEDA just scaler the workload you want, but you need to define it by yourself, you can define your Deployment with that association in it's needed, and then deploy a ScaledObject to scale that Deployment, KEDA will generate an HPA and your Deployment will be scaled, but KEDA doesn't know anything about the associations of your deployment. I mean, KEDA won't access to the node and do anything, KEDA just generates the HPA and the metrics for that HPA

[JorTurFer]

WRT adding the parameter, it's totally feasible, if that parameter can be used for filtering we can introduce it :)
Are you willing to contribute with it?

[gifflet]

No no, I'm not saying that the way isn't that, I'm just saying that KEDA doesn't manage it. KEDA just scaler the workload you want, but you need to define it by yourself, you can define your Deployment with that association in it's needed, and then deploy a ScaledObject to scale that Deployment, KEDA will generate an HPA and your Deployment will be scaled, but KEDA doesn't know anything about the associations of your deployment. I mean, KEDA won't access to the node and do anything, KEDA just generates the HPA and the metrics for that HPA

Ok, I understand.

Are you willing to contribute with it?

Yes, I intend to be contributing to this in the future. I won't start right away because this activity is not a priority at the moment and therefore I ask you to keep this issue open.

[gifflet]

Selenium grid scaler updated in #4105 and docs in kedacore/keda-docs#1039

[gifflet]

Can this issue be closed now?