Closed brtnshrdr closed 7 years ago
@brtnshrdr
This can accomplish this in two ways:
livenessProbe:
tcpSocket:
port: <kanali port>
initialDelaySeconds: 180
readinessProbe:
tcpSocket:
port: <kanali port>
initialDelaySeconds: 10
kind: ConfigMap
apiVersion: v1
metadata:
name: kanali-alive
namespace: default
data:
response: |-
[{
"route": "/alive",
"code": 200,
"method": "GET",
"body": {}
}]
Make sure mock responses are enabled
[proxy]
enable_mock_responses = true
The tcp socket method should work great. Thanks!
It would be nice if Kanali had a default health check available (and configured in the kube service files). When trying to create an Ingress with Google Container Engine, the ingress-gce process sets up an automatic health check that says a
healthy
pod is based on either the liveness probe or a request to the service with a path of/
must return a 200. If we setup Kanali on a new cluster, then create the LoadBalancer, the health check fails unless we were to deploy anApiProxy
at the/
path.