Merge pull request #59678 from Cynerva/gkk/kube-proxy-node-name

Automatic merge from submit-queue. If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>. juju: Fix Services unreachable when externalTrafficPolicy=Local on AWS **What this PR does / why we need it**: This fixes Services being unreachable when externalTrafficPolicy=Local in Juju based AWS deployments. The root cause is that kube-proxy is unable to identify local endpoints because it's using the incorrect node name. The fix is to specify node name via --hostname-override. **Which issue(s) this PR fixes**: https://github.com/juju-solutions/bundle-canonical-kubernetes/issues/489 **Special notes for your reviewer**: **Release note**: ```release-note NONE ```

Merge pull request #59678 from Cynerva/gkk/kube-proxy-node-name
52a60543 · Kubernetes Submit Queue · GitHub · 28b4bb0f · fba0119e · 52a60543
Unverified Commit 52a60543 authored Feb 10, 2018 by Kubernetes Submit Queue Committed by GitHub Feb 10, 2018
Hide whitespace changes
Inline Side-by-side

Showing with 8 additions and 38 deletions

kubernetes_worker.py ...ju/layers/kubernetes-worker/reactive/kubernetes_worker.py +8 -38

No files found.
--- a/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py
+++ b/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py
@@ -24,7 +24,7 @@ import time
 from shlex import split
 from subprocess import check_call, check_output
 from subprocess import CalledProcessError
-from socket import gethostname
+from socket import gethostname, getfqdn
 from charms import layer
 from charms.layer import snap
@@ -614,6 +614,7 @@ def configure_kube_proxy(api_servers, cluster_cidr):
    kube_proxy_opts['logtostderr'] = 'true'
    kube_proxy_opts['v'] = '0'
    kube_proxy_opts['master'] = random.choice(api_servers)
+    kube_proxy_opts['hostname-override'] = get_node_name()
    if b'lxc' in check_output('virt-what', shell=True):
        kube_proxy_opts['conntrack-max-per-core'] = '0'
@@ -989,44 +990,13 @@ def _systemctl_is_active(application):
        return False
-class GetNodeNameFailed(Exception):
-    pass
 def get_node_name():
-    # Get all the nodes in the cluster
+    kubelet_extra_args = parse_extra_args('kubelet-extra-args')
-    cmd = 'kubectl --kubeconfig={} get no -o=json'.format(kubeconfig_path)
+    cloud_provider = kubelet_extra_args.get('cloud-provider', '')
-    cmd = cmd.split()
+    if cloud_provider == 'aws':
-    deadline = time.time() + 180
+        return getfqdn()
-    while time.time() < deadline:
+    else:
-        try:
+        return gethostname()
-            raw = check_output(cmd)
-        except CalledProcessError:
-            hookenv.log('Failed to get node name for node %s.'
-                        ' Will retry.' % (gethostname()))
-            time.sleep(1)
-            continue
-        result = json.loads(raw.decode('utf-8'))
-        if 'items' in result:
-            for node in result['items']:
-                if 'status' not in node:
-                    continue
-                if 'addresses' not in node['status']:
-                    continue
-                # find the hostname
-                for address in node['status']['addresses']:
-                    if address['type'] == 'Hostname':
-                        if address['address'] == gethostname():
-                            return node['metadata']['name']
-                        # if we didn't match, just bail to the next node
-                        break
-        time.sleep(1)
-    msg = 'Failed to get node name for node %s' % gethostname()
-    raise GetNodeNameFailed(msg)
 class ApplyNodeLabelFailed(Exception):