[🐛 Bug]: Autoscaling issues when specifying most capabilities

[grant-mccarriagher]

What happened?

This issue seems distinct from the other KEDA issue that's currently open, but let me know if it's related and if this one should be closed.

We're trying to roll out Selenium Grid for k8s, and running into some autoscaling issues. When specifying just browserName, things work as expected. The appropriate number of nodes are autoscaled up, and then brought back down. However, specifying a platformName other than windows results in a session queue with no nodes being spun up. Here's a couple different scenarios we've noticed with different capabilities:

1. { browserName: "chrome", acceptInsecureCerts: true, 'goog:chromeOptions':{ ...various chrome options }, pageLoadStrategy: "eager" }: Chrome nodes are spun up appropriately, the nodes default to Windows, sessions make it to the nodes
2. { browserName: "firefox" }: Firefox nodes spun up appropriately, defaults to Windows, sessions make it to the nodes
3. { browserName: "chrome", platformName: "linux" }: Nodes do not get spun up, sessions remain in queue indefinitely
4. { browserName: "chrome", "se:downloadsEnabled": true }: The correct number of nodes are spun up, but sessions are not assigned to them and stay in the queue indefinitely. The nodes appear to lack the se:downloadsEnabled capability.

We have our own instance of KEDA that we're using which is on version 2.16.1. We're using the latest version of the chart as of this ticket, which is 0.38.4. Please let me know what additional info I can add, or if there's any questions.

Command used to start Selenium Grid with Docker (or Kubernetes)

values.yml:

global:
  seleniumGrid:
    structuredLogs: true

ingress:
  hostname: selenium-grid.internal.internal-development.our-website.com
  className: internal
  annotations:
    nginx.ingress.kubernetes.io/use-regex: "true"
    nginx.ingress.kubernetes.io/rewrite-target: /$2
    nginx.ingress.kubernetes.io/app-root: &gridAppRoot "/selenium"
  paths:
  - path: /selenium(/|$)(.*)
    pathType: ImplementationSpecific
    backend:
      service:
        name: '{{ ternary (include "seleniumGrid.router.fullname" $ ) (include "seleniumGrid.hub.fullname" $ ) $.Values.isolateComponents }}'
        port:
          number: 4444

basicAuth:
  enabled: false

autoscaling:
  enableWithExistingKEDA: true
  patchObjectFinalizers:
    serviceAccount: selenium-grid-selenium-sa
  scalingType: deployment
  scaledOptions:
    minReplicaCount: 0
    maxReplicaCount: 128

hub:
  subPath: *gridAppRoot

components:
  subPath: *gridAppRoot

Relevant log output

I don't see any errors or warnings in the logs, just that the nodes are being brought up (when platformName isn't specified).

Operating System

Linux

Docker Selenium version (image tag)

4.27.0-d6e718d

Selenium Grid chart version (chart version)

0.38.4

[github-actions]

@grant-mccarriagher, thank you for creating this issue. We will troubleshoot it as soon as we can.

---

Info for maintainers

Triage this issue by using labels.

If information is missing, add a helpful comment and then I-issue-template label.

If the issue is a question, add the I-question label.

If the issue is valid but there is no time to troubleshoot it, consider adding the help wanted label.

If the issue requires changes or fixes from an external project (e.g., ChromeDriver, GeckoDriver, MSEdgeDriver, W3C), add the applicable G-* label, and it will provide the correct link and auto-close the issue.

After troubleshooting the issue, please add the R-awaiting answer label.

Thank you!

[VietND96]

• On issue point 1 & 2, when platformName is empty, it gets converted to windows. It will be fixed in Grid core [grid]: platformName is empty should be considered as enum ANY instead of WINDOWS selenium#15036 (release in 4.28) and KEDA scaler will also be updated to cover case any fix: Selenium Grid scaler logic on platformName is set empty or any kedacore/keda#6477
• On issue point 3. the request { browserName: "chrome", platformName: "linux" }. This requires the scaler metadata also set platformName: linux (Helm config hpa in each node) to spawn up that Node stereotype. You also can see Helm config crossBrowsers to define multiple scaledObject with different metadata. Checkout example: https://github.com/SeleniumHQ/docker-selenium/blob/trunk/charts/selenium-grid/multiple-nodes-platform.yaml
• On issue point 4: { browserName: "chrome", "se:downloadsEnabled": true }, this is a special case, request assigned to Node by Grid core, the feature manage-downloads is enabled by env var SE_NODE_ENABLE_MANAGED_DOWNLOADS. We will try to expand the capabilities support in scaler metadata later.

[grant-mccarriagher]

@VietND96 Thanks for explaining this, I had no idea we needed to specify which capability combinations we support in the Values file. Everything seems to be working great now.