mlflow chart

Tested in kind on my local laptop. To test ingress in hosted kubernetes solution, we need to apply the correct ingressClass and annotations.

kubectl create ns mmm
helm upgrade --install mmm ./mlflow -f values.example.yaml --namespace mmm
  1. How can you make sure this application does not go out of hand and consume all the cluster resources?

by setting the resources constraints in the values.yaml file

  1. How would you collect logs from this app for later analysis? Where and how would you save it?

This is really a cost question. If budget allows, I would use datadog and collect all logs from the cluster globally: https://docs.datadoghq.com/containers/kubernetes/log/?tab=datadogoperator

Another option I think is solid (never actually tried to work with) is Grafana's Loki: https://grafana.com/blog/2023/04/12/how-to-collect-and-query-kubernetes-logs-with-grafana-loki-grafana-and-grafana-agent/