Merge ~paulgear/influxdb-charm/+git/influxdb-charm:master into influxdb-charm:master

[Paul Gear]

Add a basic Nagios check for InfluxDB RT#106501

[Haw Loeung (hloeung)]

LGTM, +1

review: Approve

diff --git a/files/check_influxdb.py b/files/check_influxdb.py
new file mode 100755
index 0000000..5ab579b
--- /dev/null
+++ b/files/check_influxdb.py
@@ -0,0 +1,24 @@
+#!/usr/bin/python3
+
+# Copyright (c) 2017 Canonical Ltd
+# Author: Paul Gear
+# Description: Run influxdb 'show diagnostics' command to ensure influxdb is running and operational
+
+import json
+import subprocess
+import sys
+import traceback
+
+try:
+    output = subprocess.check_output(['influx', '-execute', 'show diagnostics', '-format', 'json'])
+    j = json.loads(output.decode('utf-8'))
+    for o in j['results'][0]['series']:
+        if o['name'] == 'system' and o['columns'][-1] == 'uptime':
+            print('OK: InfluxDB uptime is {}'.format(o['values'][0][-1]))
+            sys.exit(0)
+    print('CRITICAL: Unable to determine InfluxDB uptime')
+    sys.exit(2)
+except Exception as e:
+    print('CRITICAL: Exception while attempting to check InfluxDB')
+    traceback.print_exc()
+    sys.exit(2)
diff --git a/layer.yaml b/layer.yaml
index 37d19f1..7a1dd2b 100644
--- a/layer.yaml
+++ b/layer.yaml
@@ -2,6 +2,7 @@ repo: https://git.launchpad.net/influxdb-charm
 includes:
   - layer:basic
   - layer:apt
+  - layer:nagios
   - interface:http
   - interface:grafana-source
   - interface:influxdb-api
diff --git a/reactive/influxdb.py b/reactive/influxdb.py
index 8895dd0..d7698cf 100644
--- a/reactive/influxdb.py
+++ b/reactive/influxdb.py
@@ -1,11 +1,16 @@
-from charmhelpers.core import hookenv
+from charmhelpers.contrib.charmsupport import nrpe
 from charmhelpers.core.host import service_start, service_stop
+from charmhelpers.core import hookenv
 from charmhelpers.core.templating import render
 from charms.reactive import hook, set_state, when, when_any, when_not
 
 import os
+import shutil
 
 CRONFILE = '/etc/cron.daily/influxdb-charm-backup'
+CHECK_SRC = 'files/check_influxdb.py'
+CHECK_DIR = '/usr/local/lib/nagios/plugins'
+CHECK_CMD = CHECK_DIR + '/check_influxdb.py'
 
 
 @when('admin.available')
@@ -74,6 +79,9 @@ def upgrade_charm():
     set_state('config.changed.backup_dir')
     set_state('config.changed.bind_port')
     set_state('config.changed.ip_address')
+    # If we have an NRPE relation, force run of that as well
+    if hookenv.relation_ids('nrpe-exteral-master'):
+        set_state('nrpe-external-master.available')
 
 
 @when('query.api.available')
@@ -88,3 +96,20 @@ def configure_grafana(grafana):
     config = hookenv.config()
     port = config.get('bind_port')
     grafana.provide('influxdb', port, 'Juju generated source')
+
+
+@when('nrpe-external-master.available')
+def update_nrpe_config(svc):
+    # copy the check into place
+    os.makedirs(CHECK_DIR, exist_ok=True)
+    shutil.copy(CHECK_SRC, CHECK_CMD)
+    os.chmod(CHECK_CMD, 0o755)
+    # install the nrpe configuration
+    hostname = nrpe.get_nagios_hostname()
+    nrpe_setup = nrpe.NRPE(hostname=hostname)
+    nrpe_setup.add_check(
+        shortname='influxdb',
+        description='Check InfluxDB',
+        check_cmd=CHECK_CMD,
+    )
+    nrpe_setup.write()