copr: nagios check for Copr's CDN

Relates: https://pagure.io/fedora-infrastructure/issue/10456

roles/copr/frontend/tasks/main.yml

@@ -172,3 +172,25 @@
     line: 'OPTIONS=--collector.textfile.directory /var/lib/node_exporter/textfile_collector --collector.systemd'
     backrefs: yes
   notify: restart node_exporter
+
+- name: install nrpe
+  dnf: name=nrpe state=present
+
+- name: install nrpe checks
+  template: src=copr_frontend_nrpe.cfg
+            dest=/etc/nrpe.d/copr_frontend_nrpe.cfg
+  notify:
+  - restart nrpe
+  tags:
+  - nagios_client
+  - copr_cdn
+
+- name: install the check script
+  template:
+    dest: "/usr/bin/copr-cdn-check.py"
+    src: "copr-cdn-check.py.j2"
+    group: "nagios"
+    mode: 0750
+  vars:
+    api_key: "{{ copr_cdn_api_key }}"
+  tags: copr_cdn

roles/copr/frontend/templates/copr-cdn-check.py.j2

@@ -0,0 +1,92 @@
+import sys
+import requests
+import logging
+
+# NAGIOS exit codes
+NAG_UNKNOWN = 3
+NAG_CRITICAL = 2
+NAG_WARNING = 1
+NAG_OK = 0
+
+# UptimeRobot statuses
+STATUS = (
+    'PAUSED',  # 0
+    'NOT_CHECKED_YET',  # 1
+    'UP',  # 2
+    'NOT_USED',  # 3
+    'NOT_USED',  # 4
+    'NOT_USED',  # 5
+    'NOT_USED',  # 6
+    'NOT_USED',  # 7
+    'SEEMS_DOWN',  # 8
+    'DOWN'  # 9
+)
+
+# Read only API key
+API_KEY = "{{ api_key }}"
+
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(message)s',
+    handlers=[logging.StreamHandler(sys.stdout)],
+)
+LOG = logging.getLogger()
+
+
+class UptimeRobot:
+
+    def __init__(self, api_key):
+        self.api_key = api_key
+
+    def post(self, method='getMonitors'):
+        """Send the post request to UptimeRobot api"""
+        url = "https://api.uptimerobot.com/v2/{method}".format(method=method)
+        payload = "api_key={api_key}&format=json".format(api_key=self.api_key)
+        headers = {
+            'content-type': "application/x-www-form-urlencoded",
+            'cache-control': "no-cache"
+        }
+        response = requests.request("POST", url, data=payload, headers=headers)
+        return response.json()
+
+    def get_monitor(self, name):
+        """Return specific monitor"""
+        monitors = self.post('getMonitors')
+        if 'monitors' not in monitors:
+            return None
+
+        for monitor in monitors['monitors']:
+            if monitor['friendly_name'] == name:
+                return {'name': name,
+                        'print_status': STATUS[monitor['status']],
+                        'status': monitor['status'],
+                        'url': monitor['url']
+                        }
+        return None
+
+
+def main():
+    monitor_name = "Copr's CDN"
+    uptime_robot = UptimeRobot(API_KEY)
+
+    monitor = uptime_robot.get_monitor(monitor_name)
+    if monitor:
+        result = "{name} ({url}) is {status}".format(name=monitor['name'], url=monitor['url'],
+                                                     status=monitor['print_status'])
+        if monitor['status'] in (0, 1):
+            LOG.warning(result)
+            sys.exit(NAG_WARNING)
+        elif monitor['status'] in (8, 9):
+            LOG.critical(result)
+            sys.exit(NAG_CRITICAL)
+        elif monitor['status'] == 2:
+            LOG.info(result)
+            sys.exit(NAG_OK)
+        else:
+            LOG.critical(result)
+            sys.exit(NAG_UNKNOWN)
+
+
+if __name__ == "__main__":
+    main()
+    sys.exit(NAG_UNKNOWN)

roles/copr/frontend/templates/copr_frontend_nrpe.cfg

@@ -0,0 +1 @@
+command[check_copr_cdn]=/usr/bin/copr-cdn-check.py

roles/nagios_server/files/nagios/services/copr.cfg

@@ -32,3 +32,10 @@ define service {
     check_command          check_by_nrpe!check_copr_ping
     use                    defaulttemplate
 }
+
+define service {
+    hostgroup_name         copr_front_aws, copr_front_dev_aws
+    service_description    The copr cdn status
+    check_command          check_by_nrpe!check_copr_cdn
+    use                    defaulttemplate
+}

roles/nagios_server/templates/nrpe/nrpe.cfg.j2

@@ -320,6 +320,7 @@ command[check_lock_file_age]=/usr/lib64/nagios/plugins/check_lock_file_age -w 1
 command[check_nagios]=/usr/lib64/nagios/plugins/check_nagios -e 5 -F /var/spool/nagios/status.dat -C /usr/sbin/nagios
 command[check_auditd]=/usr/lib64/nagios/plugins/check_procs -c 1:1 -C 'auditd' -u root
 command[check_copr_ping]=/usr/bin/copr-ping-check.py
+command[check_copr_cdn]=/usr/bin/copr-cdn-check.py
 command[check_cron]=/usr/lib64/nagios/plugins/check_procs -c 1:10 -C 'crond' -u root
 command[check_varnish_proc]=/usr/lib64/nagios/plugins/check_procs -c 1:2 -C 'varnishd' -u varnish
 command[check_rsyslogd_proc]=/usr/lib64/nagios/plugins/check_procs -c 1:2 -C 'rsyslogd'