KUBE_ENABLE_NODE_LOGGING and KUBE_LOGGING_DESTINATION env vars are used by kube-up.sh script. I dont know much about aws cli tool you have mentioned, but looking at the code, it doesnt look like those env vars affect that cli.

http://kubernetes.io/docs/getting-started-guides/aws/ details the steps required to bring up a kubernetes cluster on AWS using the kube-up script.

So it seems that there is no support for this on kube-aws currently, quoting one of the authors:

We are currently working on a kube-was distribution for this approach that includes Kibana for visualizing the elastic search data.

Also a suggested workaround appears in this issue page including extra details regarding it's status: https://github.com/coreos/coreos-kubernetes/issues/320

I've managed to get the cluster-level logging running on a small testing cluster started through the CoreOS kube-aws tool using the following steps. Please be aware that although I've had this running, I haven't really played with it sufficiently to be able to guarantee that all works correctly!

Enable log collection on nodes

You'll need to edit the cloud-config-worker and cloud-config-controller to export kubelet-collected logs and create the log directory

[Service] Environment="RKT_OPTS=--volume var-log,kind=host,source=/var/log --mount volume=var-log,target=/var/log" Environment=KUBELET_VERSION=v1.2.4_coreos.1 ExecStartPre=/usr/bin/mkdir -p /var/log/containers ExecStart=/usr/lib/coreos/kubelet-wrapper \ --api-servers=http://127.0.0.1:8080 \ --config=/etc/kubernetes/manifests ...other flags...

(taken from the 'Use the cluster logging add-on' section here)

Install the logging components I used the components from here (as you've already attempted). As you noticed, this does not run fluentd, and assumes that it is run as part of the cluster bootstrapping. To get fluentd running I've extracted the fluentd Daemonset definition discussed here into a separate file:

{ "apiVersion": "extensions\/v1beta1", "kind": "DaemonSet", "metadata": { "name": "fluent-elasticsearch", "namespace": "kube-system", "labels": { "k8s-app": "fluentd-logging" } }, "spec": { "template": { "metadata": { "name": "fluentd-elasticsearch", "namespace": "kube-system", "labels": { "k8s-app": "fluentd-logging" } }, "spec": { "containers": [ { "name": "fluentd-elasticsearch", "image": "gcr.io\/google_containers\/fluentd-elasticsearch:1.15", "resources": { "limits": { "memory": "200Mi" }, "requests": { "cpu": "100m", "memory": "200Mi" } }, "volumeMounts": [ { "name": "varlog", "mountPath": "\/var\/log" }, { "name": "varlibdockercontainers", "mountPath": "\/var\/lib\/docker\/containers", "readOnly": true } ] } ], "terminationGracePeriodSeconds": 30, "volumes": [ { "name": "varlog", "hostPath": { "path": "\/var\/log" } }, { "name": "varlibdockercontainers", "hostPath": { "path": "\/var\/lib\/docker\/containers" } } ] } } } }

this Daemonset runs fluentd on each of the cluster nodes.

(NOTE: Whilst I've only tried adding these components after the cluster is already running, there's no reason you shouldn't be able to add these to to the cloud-config-controller in order to bring these up at the same time the cluster is started - which is more inline with that discussed on the referenced issue)

These instruction all assume that you're working with a cluster that you're happy to restart, or haven't yet started, in order to get the logging running - which I assume from your question is the situation you're in. I've also been able to get this working on a pre-existing cluster, by manually editing the AWS settings, and can add additional information on doing this if that is in fact what you are trying to do.