lemon-apple-73935
05/11/2026, 2:13 PMkubectl, I can inspect the pods and resources without issues. What’s unclear to me is how to customize Helm values after deployment.
For example, I’d like to set:
RUNTIME_CLASS: "sysbox-runc"
as shown here:
https://github.com/All-Hands-AI/OpenHands-Cloud/blob/9f05a50b819caa553752efeb0354c69619de3bc1/charts/runtime-api/values.yaml#L53
I struggle to understand how changes like this are meant to be applied in the deployed environment.
I’m familiar with Kubernetes and Helm, but the admin UI is breaking my usual mental model of how the system is managed. Normally I’d expect some helm upgrade / helm template + kubectl apply workflow, but I can’t tell:
• where the Helm releases are managed,
• whether the UI writes back to Helm values,
• or how manual configuration changes should be applied and persisted.
Any hints are much appreciated.
Thanks!wooden-umbrella-84098
05/11/2026, 2:29 PMhelm upgrade loop isn't the right entry point here; KOTS is what reconciles state onto the cluster.
How it actually works under the hood:
- KOTS stores your config (the values you set in the admin UI) as a versioned application config.
- On deploy, KOTS renders the Helm charts with those values plus its own templating, and applies the result to the cluster.
- Any change you make in the admin UI creates a new version that you then deploy — that's the equivalent of helm upgrade in this world.
- If you helm upgrade or kubectl apply directly, KOTS will overwrite your changes on the next deploy, so don't go that route.
For the specific case of RUNTIME_CLASS: "sysbox-runc" on the runtime-api chart: that value isn't exposed as a first-class field in the admin UI today. The way to set arbitrary Helm values that we don't surface directly is through the Config screen in the KOTS admin console — there's a section for advanced/custom Helm values that gets merged into the chart values at render time. Set it there, save, and then deploy the new version from the Version history tab. After it deploys you can confirm with kubectl get pods -n <ns> -o yaml that the runtimeClassName landed where you expect.
A couple of things worth flagging:
- sysbox-runc requires the Sysbox runtime to actually be installed on the node and registered as a RuntimeClass in the cluster. The embedded installer runs k0s on the EC2 instance and doesn't ship Sysbox by default, so you'll need to install it on the host and create the RuntimeClass resource before the value will do anything useful. Happy to dig into that with you if it's not already set up.
- If the value you want to override isn't reachable from the admin UI's custom values section, let me know which one and we can either expose it or give you a workaround.
Let me know how it goes!dry-book-58714
05/11/2026, 2:36 PMdry-book-58714
05/11/2026, 2:37 PMdry-book-58714
05/11/2026, 2:38 PMdry-book-58714
05/11/2026, 2:44 PMlemon-apple-73935
05/11/2026, 2:46 PMsysbox release channel? It seems like this would take care of all the details @wooden-umbrella-84098 mentioned above which is great. However I’m not ready for such disruptive actions as I just share the OH instance with a few folks to tinker. Maybe later this week. What is the process to get onto the channel, open a support ticket via https://install.r9.all-hands.dev/openhands/support?dry-book-58714
05/11/2026, 3:04 PMsysbox release channel would be the easiest. It is in beta as it is a tricky setup with additional install scripts. I think you can ping us in this thread when you are ready! CC @average-processor-75195 on the support ticket processlemon-apple-73935
05/20/2026, 7:55 AMthere’s a section for advanced/custom Helm values that gets merged into the chart values at render time.