GCE: Services (LoadBalancer) with static ip causes panic in 1.7 #48848
Labels
area/platform/gce
kind/bug
Categorizes issue or PR as related to a bug.
sig/network
Categorizes an issue or PR as relevant to SIG Network.
Milestone
/kind bug
What happened:
Services of type LoadBalancer (GCE) with IP addresses promoted to static (by user or by orphan) cause the service controller to panic. The constant panicking will cause a reboot loop of the pod.
How to reproduce it (as minimally and precisely as possible):
Environment:
1.7.0 AND 1.7.1
GCE
Symptoms
Mitigation
Option 1:
Build your own version of
release-1.7
branch for the master. Only the kube-controller-manager image needs to be updated.Option 2:
For all affected services, set the
loadBalancerIP
field to the known external IP address.Steps:
➜ gcloud compute addresses list
NAME REGION ADDRESS STATUS
a8b891421675f11e7bee242010a80000 us-central1 35.35.35.35 IN_USE
➜ kubectl get svc --all-namespaces | grep "LoadBalancer"
default my-service LoadBalancer 10.0.99.201 35.35.35.35 80:32587/TCP,443:30446/TCP 58m
loadBalancerIP
field equal to the known IP.Example
apiVersion: v1
kind: Service
metadata:
creationTimestamp: 2017-07-13T00:09:31Z
labels:
app: echo
name: my-service
namespace: default
resourceVersion: "46816"
selfLink: /api/v1/namespaces/default/services/my-service
uid: 8b891421-675f-11e7-bee2-42010a800002
spec:
clusterIP: 10.0.99.201
externalTrafficPolicy: Cluster
loadBalancerIP: 35.35.35.35 <------- Field set to status.loadBalancer.ingress.ip below.
ports:
nodePort: 32587
port: 80
protocol: TCP
targetPort: 80
nodePort: 30446
port: 443
protocol: TCP
targetPort: 443
selector:
app: echo
sessionAffinity: ClientIP
type: LoadBalancer
status:
loadBalancer:
ingress:
The text was updated successfully, but these errors were encountered: