On Thursday, June 16, 2016, insomia <[email protected]> wrote: > Thanks for your answer, but this not solved the problem. > > I added this on frontend_deployment.yml: > > rollingUpdate: > > maxUnavailable: 100% > > > After kubernetes apply -f frontend_deployment.yml: > kubectl get events > > 3m 3m 1 frontend-dev > Deployment Normal > ScalingReplicaSet {deployment-controller } Scaled up replica set > frontend-dev-426955336 to 1 > > 3m 3m 1 frontend-dev > Deployment Normal > ScalingReplicaSet {deployment-controller } Scaled down replica set > frontend-dev-305582663 to 0 > > > So, new rs scale to 1 first, then old rs scale to 0. Also, in > documentation ( > http://kubernetes.io/docs/api-reference/extensions/v1beta1/definitions/#_v1beta1_deploymentspec) > is said that maxUnavailable set by default to fixed value of 1. So, > maxUnavailable: 100% is unnecessary. > Not really, maxUnavalable 0 will make sure the pod is not killed until the other one is **ready**.
Have you tried setting it to 0 and see if you still have downtime during deployment? Also, I have this problem not only in moment of deploy (when pod is > recreated), but more longer. Sometimes service become available after > about 5 minute, or after deletion of pod manually and it recreation. > After 5 minutes of what event? A deployment? I don't know for, but aws' load balancer have long timeouts by default and if you have one replica without "maxUnavailable: 0", it will take some minutes to route them again (unless you change timeout or the maxUnavailable ) It is always possible to connect inside the cluster, though. So, to rule out such a problem in your case: when it is unreachable from the internet, is it reachable from inside the cluster? -- You received this message because you are subscribed to the Google Groups "Containers at Google" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/google-containers. For more options, visit https://groups.google.com/d/optout.
