gcloud - kubectl rolling-update having a downtime of few seconds -

i'm using google container engine cluster using kubernetes 0.20.2.

in cluster, have 1 replication controller (2 replicas) , 1 service spec type defined loadbalancer (basic setup).

everything working fine here; want roll update different image using kubectl command:

kubectl rolling-update my-rc \   --image=gcr.io/project/gcloudid:my-image-updated \   --update-period=0m 

about understood, running command should take care of having zero-downtime. unfortunately, have been doing test using curl command in loop, , still have downtime of few seconds. ideas why happening?

the --update-period flag tells kubernetes how long wait between each pod it's rolling update to. update period set 0, kubernetes update pods @ once, causing short period of unavailability while new pods start up. should set --update-period @ least long takes each of pods initialize. default value (1 minute) should fine cases if don't want have think it.