gravity
gravity copied to clipboard
gravity-site has readiness probe failed if it not leader and produces thousands of events
Describe the bug
kubectl describe pods -n=kube-system gravity-site-xxxxxx
in case if the pod is not leader, produces events
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Warning Unhealthy 86s (x25710 over 2d23h) kubelet, ip.ec2.internal Readiness probe failed: HTTP probe failed with statuscode: 503
To Reproduce
Execute for not leader kubectl describe pods -n=kube-system gravity-site-xxxxxx
Expected behavior
pods should be healthy
Logs
Environment (please complete the following information):
- OS [e.g. Redhat 7.4]:
- Gravity [e.g. 5.5.4]:gravity-site:6.3.9
- Platform [e.g. Vmware, AWS]:
Additional context
Thanks for the report! However, this is by design. In an HA cluster only a single gravity-site is ready and serves API requests. Going to close for now.
Going to reopen actually for tracking. Not sure we're gonna be able to do something about it in near future as the internals rely quite heavily on this "single leader" model at the moment, but this is a valid complaint nonetheless.
As an idea, maybe we could achieve the same effect by utilizing pod labels and service selector instead of failing the readiness probe.