Merge pull request #41351 from chuckbutler/multi-master-patch

Automatic merge from submit-queue (batch tested with PRs 40665, 41094, 41351, 41721, 41843)

Multi master patch

**What this PR does / why we need it**: Corrects a sync files issue present when running in a HA Master configuration. This PR adds logic to syncronize on first deployment for `/etc/kubernetes/serviceaccount.key` which will cause cypto verification failure if not 1:1 on each master unit. Additionally syncs basic_auth and additional files in /srv/kubernetes. 

**Which issue this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close that issue when PR gets merged)*: fixes #41019

**Special notes for your reviewer**: This requires PR #41251 as a dependency before merging. 

**Release note**:

```release-note
Juju - K8s master charm now properly keeps distributed master files in sync for an HA control plane.
```

cluster/juju/layers/kubernetes-master/layer.yaml

@@ -2,6 +2,7 @@ repo: https://github.com/kubernetes/kubernetes.git
 includes:
   - 'layer:basic'
   - 'layer:tls-client'
+  - 'layer:leadership'
   - 'layer:debug'
   - 'interface:etcd'
   - 'interface:http'

cluster/juju/layers/kubernetes-master/reactive/kubernetes_master.py

@@ -21,6 +21,8 @@ import socket
 import string
 import json
 
+import charms.leadership
+
 from shlex import split
 from subprocess import call
 from subprocess import check_call
@@ -140,34 +142,92 @@ def configure_cni(cni):
     cni.set_config(is_master=True, kubeconfig_path='')
 
 
+@when('leadership.is_leader')
 @when('kubernetes-master.components.installed')
 @when_not('authentication.setup')
-def setup_authentication():
+def setup_leader_authentication():
     '''Setup basic authentication and token access for the cluster.'''
     api_opts = FlagManager('kube-apiserver')
     controller_opts = FlagManager('kube-controller-manager')
 
-    api_opts.add('--basic-auth-file', '/srv/kubernetes/basic_auth.csv')
-    api_opts.add('--token-auth-file', '/srv/kubernetes/known_tokens.csv')
+    service_key = '/etc/kubernetes/serviceaccount.key'
+    basic_auth = '/srv/kubernetes/basic_auth.csv'
+    known_tokens = '/srv/kubernetes/known_tokens.csv'
+
+    api_opts.add('--basic-auth-file', basic_auth)
+    api_opts.add('--token-auth-file', known_tokens)
     api_opts.add('--service-cluster-ip-range', service_cidr())
     hookenv.status_set('maintenance', 'Rendering authentication templates.')
-    htaccess = '/srv/kubernetes/basic_auth.csv'
-    if not os.path.isfile(htaccess):
+    if not os.path.isfile(basic_auth):
         setup_basic_auth('admin', 'admin', 'admin')
-    known_tokens = '/srv/kubernetes/known_tokens.csv'
     if not os.path.isfile(known_tokens):
         setup_tokens(None, 'admin', 'admin')
         setup_tokens(None, 'kubelet', 'kubelet')
         setup_tokens(None, 'kube_proxy', 'kube_proxy')
     # Generate the default service account token key
     os.makedirs('/etc/kubernetes', exist_ok=True)
-    cmd = ['openssl', 'genrsa', '-out', '/etc/kubernetes/serviceaccount.key',
+
+    cmd = ['openssl', 'genrsa', '-out', service_key,
            '2048']
     check_call(cmd)
-    api_opts.add('--service-account-key-file',
-                 '/etc/kubernetes/serviceaccount.key')
-    controller_opts.add('--service-account-private-key-file',
-                        '/etc/kubernetes/serviceaccount.key')
+    api_opts.add('--service-account-key-file', service_key)
+    controller_opts.add('--service-account-private-key-file', service_key)
+
+    # read service account key for syndication
+    leader_data = {}
+    for f in [known_tokens, basic_auth, service_key]:
+        with open(f, 'r') as fp:
+            leader_data[f] = fp.read()
+
+    # this is slightly opaque, but we are sending file contents under its file
+    # path as a key.
+    # eg:
+    # {'/etc/kubernetes/serviceaccount.key': 'RSA:2471731...'}
+    charms.leadership.leader_set(leader_data)
+
+    set_state('authentication.setup')
+
+
+@when_not('leadership.is_leader')
+@when('kubernetes-master.components.installed')
+@when_not('authentication.setup')
+def setup_non_leader_authentication():
+    api_opts = FlagManager('kube-apiserver')
+    controller_opts = FlagManager('kube-controller-manager')
+
+    service_key = '/etc/kubernetes/serviceaccount.key'
+    basic_auth = '/srv/kubernetes/basic_auth.csv'
+    known_tokens = '/srv/kubernetes/known_tokens.csv'
+
+    # This races with other codepaths, and seems to require being created first
+    # This block may be extracted later, but for now seems to work as intended
+    os.makedirs('/etc/kubernetes', exist_ok=True)
+    os.makedirs('/srv/kubernetes', exist_ok=True)
+
+    hookenv.status_set('maintenance', 'Rendering authentication templates.')
+
+    # Set an array for looping logic
+    keys = [service_key, basic_auth, known_tokens]
+    for k in keys:
+        # If the path does not exist, assume we need it
+        if not os.path.exists(k):
+            # Fetch data from leadership broadcast
+            contents = charms.leadership.leader_get(k)
+            # Default to logging the warning and wait for leader data to be set
+            if contents is None:
+                msg = "Waiting on leaders crypto keys."
+                hookenv.status_set('waiting', msg)
+                hookenv.log('Missing content for file {}'.format(k))
+                return
+            # Write out the file and move on to the next item
+            with open(k, 'w+') as fp:
+                fp.write(contents)
+
+    api_opts.add('--basic-auth-file', basic_auth)
+    api_opts.add('--token-auth-file', known_tokens)
+    api_opts.add('--service-cluster-ip-range', service_cidr())
+    api_opts.add('--service-account-key-file', service_key)
+    controller_opts.add('--service-account-private-key-file', service_key)
 
     set_state('authentication.setup')
 
@@ -185,9 +245,7 @@ def idle_status():
     if not all_kube_system_pods_running():
         hookenv.status_set('waiting', 'Waiting for kube-system pods to start')
     elif hookenv.config('service-cidr') != service_cidr():
-        msg = "Cannot change {0} to {1}".format(service_cidr(),
-                                                hookenv.config('service-cidr'))
-        hookenv.log(msg, level=hookenv.WARN)
+        msg = 'WARN: cannot change service-cidr, still using ' + service_cidr()
         hookenv.status_set('active', msg)
     else:
         hookenv.status_set('active', 'Kubernetes master running.')
@@ -305,7 +363,7 @@ def start_kube_dns():
 
     context = {
         'arch': arch(),
-        # The dictionary named pillar is a construct of the k8s template files.
+        # The dictionary named 'pillar' is a construct of the k8s template file
         'pillar': {
             'dns_server': get_dns_ip(),
             'dns_replicas': 1,

cluster/juju/layers/kubernetes-master/tactics/update_addons.py

@@ -45,31 +45,46 @@ def clean_addon_dir(addon_dir):
     os.makedirs(addon_dir)
 
 
-@contextmanager
-def kubernetes_repo():
-    """ Shallow clone kubernetes repo and clean up when we are done """
-    repo = "https://github.com/kubernetes/kubernetes.git"
-    path = tempfile.mkdtemp(prefix="kubernetes")
-    try:
-        log.info("Cloning " + repo)
-        cmd = ["git", "clone", "--depth", "1", repo, path]
-        process = subprocess.Popen(cmd, stderr=subprocess.PIPE)
+def run_with_logging(command):
+    """ Run a command with controlled logging """
+    log.debug("Running: %s" % command)
+    process = subprocess.Popen(command, stderr=subprocess.PIPE)
     stderr = process.communicate()[1].rstrip()
     process.wait()
     if process.returncode != 0:
         log.error(stderr)
-            raise Exception("clone failed: exit code %d" % process.returncode)
+        raise Exception("%s: exit code %d" % (command, process.returncode))
     log.debug(stderr)
+
+
+@contextmanager
+def kubernetes_repo():
+    """ Yield a kubernetes repo to copy addons from.
+
+    If KUBE_VERSION is set, this will clone the local repo and checkout the
+    corresponding branch. Otherwise, the local branch will be used. """
+    repo = os.path.abspath("../../../..")
+    if "KUBE_VERSION" in os.environ:
+        branch = os.environ["KUBE_VERSION"]
+        log.info("Cloning %s with branch %s" % (repo, branch))
+        path = tempfile.mkdtemp(prefix="kubernetes")
+        try:
+            cmd = ["git", "clone", repo, path, "-b", branch]
+            run_with_logging(cmd)
             yield path
         finally:
             shutil.rmtree(path)
+    else:
+        log.info("Using local repo " + repo)
+        yield repo
 
 
-def add_addon(source, dest):
-    """ Add an addon manifest from the given source.
+def add_addon(repo, source, dest):
+    """ Add an addon manifest from the given repo and source.
 
     Any occurrences of 'amd64' are replaced with '{{ arch }}' so the charm can
     fill it in during deployment. """
+    source = os.path.join(repo, "cluster/addons", source)
     if os.path.isdir(dest):
         dest = os.path.join(dest, os.path.basename(source))
     log.debug("Copying: %s -> %s" % (source, dest))
@@ -86,20 +101,26 @@ def update_addons(dest):
     with kubernetes_repo() as repo:
         log.info("Copying addons to charm")
         clean_addon_dir(dest)
-        add_addon(repo + "/cluster/addons/dashboard/dashboard-controller.yaml",
-                  dest)
-        add_addon(repo + "/cluster/addons/dashboard/dashboard-service.yaml",
-                  dest)
-        add_addon(repo + "/cluster/addons/dns/kubedns-controller.yaml.in",
+        add_addon(repo, "dashboard/dashboard-controller.yaml", dest)
+        add_addon(repo, "dashboard/dashboard-service.yaml", dest)
+        try:
+            add_addon(repo, "dns/kubedns-controller.yaml.in",
                       dest + "/kubedns-controller.yaml")
-        add_addon(repo + "/cluster/addons/dns/kubedns-svc.yaml.in",
+            add_addon(repo, "dns/kubedns-svc.yaml.in",
                       dest + "/kubedns-svc.yaml")
-        influxdb = "/cluster/addons/cluster-monitoring/influxdb"
-        add_addon(repo + influxdb + "/grafana-service.yaml", dest)
-        add_addon(repo + influxdb + "/heapster-controller.yaml", dest)
-        add_addon(repo + influxdb + "/heapster-service.yaml", dest)
-        add_addon(repo + influxdb + "/influxdb-grafana-controller.yaml", dest)
-        add_addon(repo + influxdb + "/influxdb-service.yaml", dest)
+        except IOError as e:
+            # fall back to the older filenames
+            log.debug(e)
+            add_addon(repo, "dns/skydns-rc.yaml.in",
+                      dest + "/kubedns-controller.yaml")
+            add_addon(repo, "dns/skydns-svc.yaml.in",
+                      dest + "/kubedns-svc.yaml")
+        influxdb = "cluster-monitoring/influxdb"
+        add_addon(repo, influxdb + "/grafana-service.yaml", dest)
+        add_addon(repo, influxdb + "/heapster-controller.yaml", dest)
+        add_addon(repo, influxdb + "/heapster-service.yaml", dest)
+        add_addon(repo, influxdb + "/influxdb-grafana-controller.yaml", dest)
+        add_addon(repo, influxdb + "/influxdb-service.yaml", dest)
 
 # Entry points
 
@@ -151,8 +172,8 @@ def parse_args():
 def main():
     """ Update addons into the layer's templates/addons folder """
     parse_args()
-    dest = os.path.abspath(os.path.join(os.path.dirname(__file__),
-                                        "../templates/addons"))
+    os.chdir(os.path.join(os.path.dirname(__file__), ".."))
+    dest = "templates/addons"
     update_addons(dest)