Is this a bug report or a feature request?:
bug report
What happened:
A service of type loadBalancer, announced via BGP, lost all its pod because their health checks failed. The endpoint list was empty because of that, but metallb kept announcing the IP address.
For externalTrafficPolicy=Local, the announcement was withdrawn.
What you expected to happen:
With no healthy endpoints, the announcement should be withdrawn. For us, this is required in the context of a failover to another datacenter where the service is still healthy.
How to reproduce it (as minimally and precisely as possible):
Single-Pod deployment (e.g. nginx) with a service of type LoadBalancer and externalTrafficPolicy=Cluster
When scaling the deployment down to 0, the announcement stays up, but shouldn't.
Anything else we need to know?:
As mentioned the withdrawal of the announcement when the cluster does not contain a healthy endpoint anymore is essential for a redundancy failover to another datacenter.
Environment:
- MetalLB version: 0.6.2 and 0.7.3
- Kubernetes version: 1.11.2
- BGP router type/version: Cisco 9k
- OS (e.g. from /etc/os-release): ContainerLinux 1855.4.0
- Kernel (e.g.
uname -a): 4.14.67-coreos
Is this a bug report or a feature request?:
bug report
What happened:
A service of type loadBalancer, announced via BGP, lost all its pod because their health checks failed. The endpoint list was empty because of that, but metallb kept announcing the IP address.
For externalTrafficPolicy=Local, the announcement was withdrawn.
What you expected to happen:
With no healthy endpoints, the announcement should be withdrawn. For us, this is required in the context of a failover to another datacenter where the service is still healthy.
How to reproduce it (as minimally and precisely as possible):
Single-Pod deployment (e.g. nginx) with a service of type LoadBalancer and externalTrafficPolicy=Cluster
When scaling the deployment down to 0, the announcement stays up, but shouldn't.
Anything else we need to know?:
As mentioned the withdrawal of the announcement when the cluster does not contain a healthy endpoint anymore is essential for a redundancy failover to another datacenter.
Environment:
uname -a): 4.14.67-coreos