To fully benefit from running a highly available prometheus-adapter, users need to tweak the configuration of kube-apiserver to correctly load balance the requests between the instances of prometheus-adapter. This is done by passing --enable-aggregator-routing=true to kube-apiserver. If we don't pass this argument to kube-apiserver, we end up with a prometheus-adapter that is not entirely highly available and with an instance that will be wasting resources since it will never do anything. To prevent that, it would be great to add some documentation to tell the users that if they want to run the kube-prometheus stack in a highly available mode, this flag needs to be given to kube-apiserver.
To fully benefit from running a highly available prometheus-adapter, users need to tweak the configuration of kube-apiserver to correctly load balance the requests between the instances of prometheus-adapter. This is done by passing
--enable-aggregator-routing=true
to kube-apiserver. If we don't pass this argument to kube-apiserver, we end up with a prometheus-adapter that is not entirely highly available and with an instance that will be wasting resources since it will never do anything. To prevent that, it would be great to add some documentation to tell the users that if they want to run the kube-prometheus stack in a highly available mode, this flag needs to be given to kube-apiserver.