add nrpe-external-master relation to kubernetes-master and kubernetes-worker

For now, the checks are very basic and only check if the systemd
services are running properly.

cluster/juju/layers/kubeapi-load-balancer/layer.yaml

@@ -1,5 +1,6 @@
 repo: https://github.com/kubernetes/kubernetes.git
 includes:
+  - 'layer:nagios'
   - 'layer:nginx'
   - 'layer:tls-client'
   - 'interface:public-address'

cluster/juju/layers/kubeapi-load-balancer/reactive/load_balancer.py

@@ -19,8 +19,10 @@ import socket
 import subprocess
 
 from charms import layer
-from charms.reactive import when
+from charms.reactive import when, when_any, when_not
+from charms.reactive import set_state, remove_state
 from charmhelpers.core import hookenv
+from charmhelpers.contrib.charmsupport import nrpe
 
 from charms.layer import nginx
 
@@ -111,3 +113,41 @@ def provide_loadbalancing(loadbalancer):
     the subordinates can get the public address of this loadbalancer.'''
     loadbalancer.set_address_port(hookenv.unit_get('public-address'),
                                   hookenv.config('port'))
+
+
+@when('nrpe-external-master.available')
+@when_not('nrpe-external-master.initial-config')
+def initial_nrpe_config(nagios=None):
+    set_state('nrpe-external-master.initial-config')
+    update_nrpe_config(nagios)
+
+
+@when('nginx.available')
+@when('nrpe-external-master.available')
+@when_any('config.changed.nagios_context',
+          'config.changed.nagios_servicegroups')
+def update_nrpe_config(unused=None):
+    services = ('nginx',)
+
+    hostname = nrpe.get_nagios_hostname()
+    current_unit = nrpe.get_nagios_unit_name()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+    nrpe.add_init_service_checks(nrpe_setup, services, current_unit)
+    nrpe_setup.write()
+
+
+@when_not('nrpe-external-master.available')
+@when('nrpe-external-master.initial-config')
+def remove_nrpe_config(nagios=None):
+    remove_state('nrpe-external-master.initial-config')
+
+    # List of systemd services for which the checks will be removed
+    services = ('nginx',)
+
+    # The current nrpe-external-master interface doesn't handle a lot of logic,
+    # use the charm-helpers code for now.
+    hostname = nrpe.get_nagios_hostname()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+
+    for service in services:
+        nrpe_setup.remove_check(shortname=service)

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

@@ -4,11 +4,12 @@ includes:
   - 'layer:tls-client'
   - 'layer:leadership'
   - 'layer:debug'
+  - 'layer:nagios'
+  - 'interface:ceph-admin'
   - 'interface:etcd'
   - 'interface:http'
   - 'interface:kubernetes-cni'
   - 'interface:kube-dns'
-  - 'interface:ceph-admin'
   - 'interface:public-address'
 options:
   basic:

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

@@ -33,8 +33,7 @@ from charms import layer
 from charms.reactive import hook
 from charms.reactive import remove_state
 from charms.reactive import set_state
-from charms.reactive import when
+from charms.reactive import when, when_any, when_not
-from charms.reactive import when_not
 from charms.reactive.helpers import data_changed
 from charms.kubernetes.flagmanager import FlagManager
 
@@ -43,6 +42,7 @@ from charmhelpers.core import host
 from charmhelpers.core import unitdata
 from charmhelpers.core.templating import render
 from charmhelpers.fetch import apt_install
+from charmhelpers.contrib.charmsupport import nrpe
 
 
 dashboard_templates = [
@@ -489,6 +489,44 @@ def ceph_storage(ceph_admin):
     set_state('ceph-storage.configured')
 
 
+@when('nrpe-external-master.available')
+@when_not('nrpe-external-master.initial-config')
+def initial_nrpe_config(nagios=None):
+    set_state('nrpe-external-master.initial-config')
+    update_nrpe_config(nagios)
+
+
+@when('kubernetes-master.components.started')
+@when('nrpe-external-master.available')
+@when_any('config.changed.nagios_context',
+          'config.changed.nagios_servicegroups')
+def update_nrpe_config(unused=None):
+    services = ('kube-apiserver', 'kube-controller-manager', 'kube-scheduler')
+
+    hostname = nrpe.get_nagios_hostname()
+    current_unit = nrpe.get_nagios_unit_name()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+    nrpe.add_init_service_checks(nrpe_setup, services, current_unit)
+    nrpe_setup.write()
+
+
+@when_not('nrpe-external-master.available')
+@when('nrpe-external-master.initial-config')
+def remove_nrpe_config(nagios=None):
+    remove_state('nrpe-external-master.initial-config')
+
+    # List of systemd services for which the checks will be removed
+    services = ('kube-apiserver', 'kube-controller-manager', 'kube-scheduler')
+
+    # The current nrpe-external-master interface doesn't handle a lot of logic,
+    # use the charm-helpers code for now.
+    hostname = nrpe.get_nagios_hostname()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+
+    for service in services:
+        nrpe_setup.remove_check(shortname=service)
+
+
 def create_addon(template, context):
     '''Create an addon from a template'''
     source = 'addons/' + template

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

@@ -1,9 +1,10 @@
 repo: https://github.com/kubernetes/kubernetes.git
 includes:
   - 'layer:basic'
-  - 'layer:docker'
-  - 'layer:tls-client'
   - 'layer:debug'
+  - 'layer:docker'
+  - 'layer:nagios'
+  - 'layer:tls-client'
   - 'interface:http'
   - 'interface:kubernetes-cni'
   - 'interface:kube-dns'

cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py

@@ -24,13 +24,14 @@ from socket import gethostname
 from charms import layer
 from charms.reactive import hook
 from charms.reactive import set_state, remove_state
-from charms.reactive import when, when_not
+from charms.reactive import when, when_any, when_not
 from charms.reactive.helpers import data_changed
 from charms.kubernetes.flagmanager import FlagManager
 from charms.templating.jinja2 import render
 
 from charmhelpers.core import hookenv
 from charmhelpers.core.host import service_stop
+from charmhelpers.contrib.charmsupport import nrpe
 
 
 kubeconfig_path = '/srv/kubernetes/config'
@@ -447,6 +448,44 @@ def kubectl_manifest(operation, manifest):
         return kubectl_success(operation, '-f', manifest)
 
 
+@when('nrpe-external-master.available')
+@when_not('nrpe-external-master.initial-config')
+def initial_nrpe_config(nagios=None):
+    set_state('nrpe-external-master.initial-config')
+    update_nrpe_config(nagios)
+
+
+@when('kubernetes-worker.config.created')
+@when('nrpe-external-master.available')
+@when_any('config.changed.nagios_context',
+          'config.changed.nagios_servicegroups')
+def update_nrpe_config(unused=None):
+    services = ('kubelet', 'kube-proxy')
+
+    hostname = nrpe.get_nagios_hostname()
+    current_unit = nrpe.get_nagios_unit_name()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+    nrpe.add_init_service_checks(nrpe_setup, services, current_unit)
+    nrpe_setup.write()
+
+
+@when_not('nrpe-external-master.available')
+@when('nrpe-external-master.initial-config')
+def remove_nrpe_config(nagios=None):
+    remove_state('nrpe-external-master.initial-config')
+
+    # List of systemd services for which the checks will be removed
+    services = ('kubelet', 'kube-proxy')
+
+    # The current nrpe-external-master interface doesn't handle a lot of logic,
+    # use the charm-helpers code for now.
+    hostname = nrpe.get_nagios_hostname()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+
+    for service in services:
+        nrpe_setup.remove_check(shortname=service)
+
+
 def _systemctl_is_active(application):
     ''' Poll systemctl to determine if the application is running '''
     cmd = ['systemctl', 'is-active', application]