Hi everyone! I'm migrating to the 1.2 service definition. That is a real improvement! But I have a question. Context: I'm working on distributing the load on our own kubernetes cluster better. We are building and containerizing bento's and then deploying them as a Deployment of pods on our cluster. Ideally, I would like to use a liveness probe so that the kubernetes Service automatically redirects the incoming requests to available pods. Ideally a pod fails the liveness probe if it is already processing its (max) concurrent requests. I cannot figure out how to do it though. Question: How can I configure a bento so that I can use an endpoint as a liveness probe that fails when it is processing more than the (max) concurrent requests? What I've tried: • Reducing the (max) concurrent requests. The /livez endpoint however never fails with 419 Too Many Requests. It sometimes takes longer to respond, but it is always successfull. • Try to access the

requests_in_progress

gauge, and use it to create a custom endpoint that fails when the value is above some threshold. I could not find a way to get that gauge because it is implemented in the middleware. Also any custom endpoint will have the POST method, and a liveness probe needs the GET method. Would love to have some suggestions on how to do this. Using BentoCloud is definitely on our roadmap, but not for now.