Nick Cardin
Nick Cardin
Until we get this working, we should display a better error message indication that direct k8 access is required for this.
Should also handle `kubectl` not being installed
Related to https://github.com/infinyon/fluvio/issues/1245
Need to test for stability
Scope and split into tasks
Approach could be to consume latest record and use it's timestamp. Need to double check that record timestamp is accurate
Maybe `--cleanup-cluster-on-error` should be the default
Suggest `fluvio cluster delete` for clean install or `fluvio cluster start --upgrade` for upgrade
On Linux using `kubectl create` against minikube, the max that is allows is about 1.5M, anything larger and the server responds with "etcdserver: request is too large". So this may...
Maybe for stop we can just set `replicas` to 0.