keepalived-vip v0.30 cannot be aware of the backend pod scaling

[panpan0000]

Summary:

directlly pull the image (aledbf/kube-keepalived-vip)
v0.29 was good, but when >=v0.30 , when the pod of service being scaled, then the /etc/keepalived/keepalived.conf will not get updated, until next time configmap changed .

HOW To Reproduce:

1.deploy v0.30/v0.32 keepalived-vip daemonset
2.add configmap like 10.6.111.50: default/myapp
3. kubectl scale --replicas=2 deploy myapp , from 1 pod to 2 pods
4. kubectl exec -it kube-keepalived-vip-xxx cat /etc/keepalived/keepalived.conf
5. found that real_server remains single item , at least for 1 mins waiting.
6. delete the daemonset, and update yaml to using v0.29 image, the keepalived.conf will get changed in 3-5 secs when app scaled.

@aledbf

[panpan0000]

I believe it's because of introducing of dumb-init, so the SIGHUP could not be propagated to keepalived process, but I will confirm very soon

[aledbf]

@panpan0000 please test aledbf/kube-keepalived-vip:0.33

[aledbf]

Closing. Fixed in master. Please use 0.33

[panpan0000]

wo~ how could you be so fast !!!!!
I'm working to root cause still.....aha

[aledbf]

@panpan0000 this was the issue c9ac529
We always need to render the template and compare the md5 of the running configuration and the new one

[panpan0000]

Yes, I confirm v0.33 works well !
now understand, bug introduced in e37095b
Thank you man, you made my day