GitHub user parshimers added a comment to the discussion: Proposal - Support Virtual Machine runtimes for compute units
I think this is a great idea, and it's amazing that it can be done without a lot of invasive infrastructure change due to how k8s works. I agree that we should recommend to prospective maintainers of a production Texera deployment that it is highly recommended to use Kata or some other CRI that is backed by VMs ultimately. Anything less is a half-measure, because users can put literally anything into a workflow, by design. However I think the way it should be exposed in the Helm charts is basically just having a way to set the runtime class for CUs. (i.e. option 1 ). For option 2 we can have a description about how to change the default runtime class, but I don't think we should strain ourselves to imagine every possible situation. As long as we give them the tools and necessary information, I think that's more than enough. GitHub link: https://github.com/apache/texera/discussions/8430#discussioncomment-18393088 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
