zalando-incubator / kube-ingress-aws-controller / 16934234606

72%

master: 72%

push

github

worker: delay ingress status update for new load balancers (#753)

Within
https://github.com/zalando-incubator/kube-ingress-aws-controller/pull/748
the controller was improved to wait until loadbalancer state becomes
"active" before updating ingress status which triggers traffic switch to
the new loadbalancer (e.g. via external-dns).

It turns out that while new loadbalancer state is active its target
groups may not have discovered targets yet.

This change delays ingress status update for a new loadbalancer to give
its target groups time to discover targets. The delay is configured by a
new `min-loadbalancer-age` flag. Default value was chosen after testing
in an AWS account.

The more elaborate alternative is to call `describe-target-health` and
check `TargetHealthDescriptions`.

Other changes:

* unexport loadbalancer state fields
* make loadbalancer state accessors its members
* skip logging of empty loadbalancer ARNs in GetStackLBStates
* use structured logging in updateIngress
* introduce AWS and Kubernetes API internal mocks using
github.com/stretchr/testify/mock

26 of 46 new or added lines in 4 files covered. (56.52%)

2861 of 3994 relevant lines covered (71.63%)

10.66 hits per line