Exporting jHiccup metrics

Setup

Download jHiccup:

wget http://www.azul.com/files/jHiccup-2.0.6-dist.tar.gz
tar xvzf jHiccup-2.0.6-dist.tar.gz

Save the following script as jhiccup_exporter, make executable, and put on PATH:

#!/usr/bin/env python
#
# Exports jHiccup statistics from jHiccup log to collectd.
#
# Getting jHiccup agent:
#
#    wget http://www.azul.com/files/jHiccup-2.0.6-dist.tar.gz
#    tar xvzf jHiccup-2.0.6-dist.tar.gz
#
# Running the Java client to be monitored:
#
#    export _JAVA_OPTIONS="$_JAVA_OPTIONS -javaagent:jHiccup-2.0.6/jHiccup.jar=\"-l hiccup.log -d 0 -i 1000\""
#    cassandra-stress …
#
# Running this exporter:
#
#    tail -f hiccup.log | ./jhiccup_exporter
#
# Exports the following metrics (in milliseconds):
#
#    jhiccup-0/gauge-hiccup_max
#

import sys
import socket
import os
import time

hostname = socket.gethostname().split('.')[0]

server = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
server.connect('/var/run/collectd-unixsock')

def readlines(sock, recv_buffer=4096, delim='\n'):
    buffer = ''
    data = True
    while data:
        data = sock.recv(recv_buffer)
        buffer += data
        while buffer.find(delim) != -1:
            line, buffer = buffer.split('\n', 1)
            yield line
    return

class ValuePack:
    def __init__(self):
        self.messages = []
        self.time = time.time()
    def add(self, plugin, plugin_instance, type, type_instance, val):
        self.messages.append('PUTVAL "%s/%s-%s/%s-%s" %f:%f\n' % (hostname, plugin, plugin_instance, type, type_instance, self.time, val))
    def send(self):
        for msg in self.messages:
            server.send(msg)
        lines = readlines(server)
        for msg in self.messages:
            reply = next(lines)
            if not reply.startswith('0 Success'):
                print('collectd error: ' + reply)

start_time = time.time()
plugin = 'jhiccup'
plugin_instance = '0'

with os.fdopen(sys.stdin.fileno(), 'r', 1) as input:
    while True:
        line = sys.stdin.readline()

        if line.startswith('#') or line.startswith('"'):
            continue

        if time.time() - start_time < 1.0: # Skip existing lines
            continue

        cols = line.split(',')

        hiccup_max = float(cols[2].rstrip(','))

        pack = ValuePack()
        pack.add(plugin, plugin_instance, 'gauge', 'hiccup_max', hiccup_max)
        pack.send()

    server.close()

collectd must be configured with the UNIX socket plugin. The script assumes that the collectd socket is at /var/run/collectd-unixsock.

Running

Run the Java client to be monitored like this:

export _JAVA_OPTIONS="$_JAVA_OPTIONS -javaagent:jHiccup-2.0.6/jHiccup.jar=\"-l hiccup.log -d 0 -i 1000\""
cassandra-stress …

Then run the exporter on the log file:

tail -f hiccup.log | jhiccup_exporter

It will export the following collectd metric (maximum delay in milliseconds):

jhiccup-0/gauge-hiccup_max

You don't have to restart the exporter when restarting cassandra-stress (that's due to tail -f).