Add Collectd as a Monitor Type

A plugin for collectd is added to use collectd on compute as a
monitor type. Monitor files are updated accordingly.
The inspector now listens on all interfaces instead of only
localhost to enable it to communicate with compute node.

JIRA: DOCTOR-86
JIRA: DOCTOR-101

Change-Id: Idc834d428152e4687020eff7d8db36a652b1bf86
Signed-off-by: Umar Farooq <umar.farooq@neclab.eu>

2 files changed, 142 insertions, 0 deletions

diff --git a/tests/lib/monitors/sample/monitor.py b/tests/lib/monitors/sample/monitor.py
new file mode 100644
index 00000000..7450c534
--- /dev/null
+++ b/tests/lib/monitors/sample/monitor.py
@@ -0,0 +1,124 @@
+##############################################################################
+# Copyright (c) 2016 NEC Corporation and others.
+#
+# All rights reserved. This program and the accompanying materials
+# are made available under the terms of the Apache License, Version 2.0
+# which accompanies this distribution, and is available at
+# http://www.apache.org/licenses/LICENSE-2.0
+##############################################################################
+
+import argparse
+from datetime import datetime
+import json
+import logger as doctor_log
+import requests
+import socket
+import time
+
+from keystoneauth1 import session
+from congressclient.v1 import client
+
+import identity_auth
+
+# NOTE: icmp message with all zero data (checksum = 0xf7ff)
+#       see https://tools.ietf.org/html/rfc792
+ICMP_ECHO_MESSAGE = '\x08\x00\xf7\xff\x00\x00\x00\x00'
+
+SUPPORTED_INSPECTOR_TYPES = ['sample', 'congress']
+
+LOG = doctor_log.Logger('doctor_monitor').getLogger()
+
+
+class DoctorMonitorSample(object):
+
+    interval = 0.1  # second
+    timeout = 0.1  # second
+    event_type = "compute.host.down"
+
+    def __init__(self, args):
+        if args.inspector_type not in SUPPORTED_INSPECTOR_TYPES:
+            raise Exception("Inspector type '%s' not supported", args.inspector_type)
+
+        self.hostname = args.hostname
+        self.inspector_type = args.inspector_type
+        self.ip_addr = args.ip or socket.gethostbyname(self.hostname)
+
+        if self.inspector_type == 'sample':
+            self.inspector_url = 'http://127.0.0.1:12345/events'
+        elif self.inspector_type == 'congress':
+            auth=identity_auth.get_identity_auth()
+            self.session=session.Session(auth=auth)
+            congress = client.Client(session=self.session, service_type='policy')
+            ds = congress.list_datasources()['results']
+            doctor_ds = next((item for item in ds if item['driver'] == 'doctor'),
+                             None)
+
+            congress_endpoint = congress.httpclient.get_endpoint(auth=auth)
+            self.inspector_url = ('%s/v1/data-sources/%s/tables/events/rows' %
+                                  (congress_endpoint, doctor_ds['id']))
+
+    def start_loop(self):
+        LOG.debug("start ping to host %(h)s (ip=%(i)s)" % {'h': self.hostname,
+                                                       'i': self.ip_addr})
+        sock = socket.socket(socket.AF_INET, socket.SOCK_RAW,
+                             socket.IPPROTO_ICMP)
+        sock.settimeout(self.timeout)
+        while True:
+            try:
+                sock.sendto(ICMP_ECHO_MESSAGE, (self.ip_addr, 0))
+                data = sock.recv(4096)
+            except socket.timeout:
+                LOG.info("doctor monitor detected at %s" % time.time())
+                self.report_error()
+                LOG.info("ping timeout, quit monitoring...")
+                return
+            time.sleep(self.interval)
+
+    def report_error(self):
+        payload = [
+            {
+                'id': 'monitor_sample_id1',
+                'time': datetime.now().isoformat(),
+                'type': self.event_type,
+                'details': {
+                    'hostname': self.hostname,
+                    'status': 'down',
+                    'monitor': 'monitor_sample',
+                    'monitor_event_id': 'monitor_sample_event1'
+                },
+            },
+        ]
+        data = json.dumps(payload)
+
+        if self.inspector_type == 'sample':
+            headers = {'content-type': 'application/json'}
+            requests.post(self.inspector_url, data=data, headers=headers)
+        elif self.inspector_type == 'congress':
+            headers = {
+                'Content-Type': 'application/json',
+                'Accept': 'application/json',
+                'X-Auth-Token':self.session.get_token(),
+            }
+            requests.put(self.inspector_url, data=data, headers=headers)
+
+
+def get_args():
+    parser = argparse.ArgumentParser(description='Doctor Sample Monitor')
+    parser.add_argument('hostname', metavar='HOSTNAME', type=str, nargs='?',
+                        help='a hostname to monitor connectivity')
+    parser.add_argument('ip', metavar='IP', type=str, nargs='?',
+                        help='an IP address to monitor connectivity')
+    parser.add_argument('inspector_type', metavar='INSPECTOR_TYPE', type=str, nargs='?',
+                        help='inspector to report',
+                        default='sample')
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    monitor = DoctorMonitorSample(args)
+    monitor.start_loop()
+
+
+if __name__ == '__main__':
+    main()
diff --git a/tests/lib/monitors/sample/sample b/tests/lib/monitors/sample/sample
new file mode 100644
index 00000000..1d310333
--- /dev/null
+++ b/tests/lib/monitors/sample/sample
@@ -0,0 +1,18 @@
+#!/bin/bash
+
+function start_monitor_sample {
+    cp $TOP_DIR/lib/monitors/sample/monitor.py $TOP_DIR/monitor.py
+    pgrep -f "python monitor.py" && return 0
+    sudo -E python monitor.py "$COMPUTE_HOST" "$COMPUTE_IP" "$INSPECTOR_TYPE" \
+        > monitor.log 2>&1 &
+}
+
+function stop_monitor_sample {
+    pgrep -f "python monitor.py" || return 0
+    sudo kill $(pgrep -f "python monitor.py")
+}
+
+function cleanup_monitor_sample {
+    rm monitor.py
+    return
+}