问题
I am using the following code in statefulset.yml
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: geth
namespace: prod
spec:
serviceName: geth-service
replicas: 2
selector:
matchLabels:
app: geth-node
template:
metadata:
labels:
app: geth-node
spec:
containers:
- name: geth-node
image: <My image>
imagePullPolicy: Always
livenessProbe:
httpGet:
path: /
port: 8545
initialDelaySeconds: 20 #wait this period after staring fist time
periodSeconds: 15 # polling interval
timeoutSeconds: 5 # wish to receive response within this time period
readinessProbe:
httpGet:
path: /
port: 8545
initialDelaySeconds: 20 #wait this period after staring fist time
periodSeconds: 15 # polling interval
timeoutSeconds: 5
and I have the following in my ingress.yml
kind: Ingress
apiVersion: extensions/v1beta1
metadata:
name: prod-ingress
namespace: prod
annotations:
kubernetes.io/ingress.class: "gce"
kubernetes.io/ingress.global-static-ip-name: "mystaticip"
spec:
tls:
- secretName: my-tls
hosts:
- myhost.com
rules:
- host: myhost
http:
paths:
- path: /rpc
backend:
serviceName: gethrpc-service
servicePort: 80
My container is exposing the port 8545
, so I configured the probe as above for my statefulset
.
When I create this statefulset and ingress, a load balancer is automatically created.
I get two Backend services ( I think two ports for two nodes - I am running it on two nodes)
However, I am confused about the following two healthchecks created for these backend services.
1. Path: /healthz Port number: 30695
2. Path: / Port number: 30190
- I don't understand why these health checks are different for different nodes.
- I configured the path to be
/
, so IMO both should have had the path/
PS: I am running a two node cluster and 2 pods, one on each node.
回答1:
The path [/healthz] is used for the default Backend for GLBC, take a look here. As you know The default backend is a service which handles all URL paths and hosts the GLBC controller doesn't understand.
The path [/] is most probably for you service.
来源:https://stackoverflow.com/questions/51869789/understanding-healthchecks-for-backend-services-on-gke-when-using-ingress