Simple Python script to store RADOS read/write latency in Graphite

ceph-rados-latency.py

#!/usr/bin/env python

'''
This script writes X MegaByte random data to 10 RADOS objects and read the data again afterwards

It writes these statistics towards a Graphite server so you can see what the latency
inside your cluster is for both Read and Write

Author: Wido den Hollander <[email protected]>
'''

import argparse
import rados
import uuid
import time
import socket

def rados_bench(ceph_conf="/etc/ceph/ceph.conf", pool="data", object_size=1, num_objects=10):
    try:
        r = rados.Rados(conffile=ceph_conf)
        r.conf_set("client_mount_timeout", "10");
        r.conf_set("rados_mon_op_timeout", "10");
        r.conf_set("rados_osd_op_timeout", "10");
        r.connect()
        ioctx = r.open_ioctx(pool)

        write_total_time = 0
        read_total_time = 0

        i = 0

        z = open('/dev/zero', 'r')
        data = z.read(object_size * 1024 * 1024)

        while i < num_objects:
            obj = str(uuid.uuid4())

            start_write = time.clock()
            ioctx.write_full(obj, data)
            write_total_time += (time.clock() - start_write) * 1000

            start_read = time.clock()
            ioctx.read(obj)
            read_total_time += (time.clock() - start_read) * 1000

            ioctx.remove_object(obj)
            i += 1

        ioctx.close()

        return {'write_time': write_total_time / num_objects, 'read_time': read_total_time / num_objects}
    except Exception as e:
        raise

def collect_metric(host, port, name, value, timestamp):
    sock = socket.create_connection((host, port))
    sock.send("%s %0.3f %d\n" % (name, value, timestamp))
    sock.close()

if __name__ == '__main__':
    parser = argparse.ArgumentParser(description='Measure Ceph RADOS latency and write to Graphite')
    parser.add_argument("--host", action="store", dest="graphite_host", default="localhost", help="The graphite host")
    parser.add_argument("--port", action="store", dest="graphite_port", type=int, default=2005, help="The graphite port")
    parser.add_argument("--prefix", action="store", dest="graphite_prefix", default="ceph", help="The prefix in graphite")
    parser.add_argument("--size", action="store", dest="object_size", type=int, default=1, help="The object size in MegaBytes")
    parser.add_argument("--num", action="store", dest="num_objects", type=int, default=10, help="The number of objects to write and read")
    parser.add_argument("--pool", action="store", dest="rados_pool", default="data", help="The RADOS pool to write in")
    parser.add_argument("--conf", action="store", dest="ceph_conf", default="/etc/ceph/ceph.conf", help="The ceph configuration file")
    conf = parser.parse_args()

    now = int(time.time())
    try:
        stats = rados_bench(ceph_conf=conf.ceph_conf, pool=conf.rados_pool, object_size=conf.object_size, num_objects=conf.num_objects)
    except Exception as e:
        stats = {'write_time': 1000, 'read_time': 1000}
        pass

    try:
        collect_metric(conf.graphite_host, conf.graphite_port, conf.graphite_prefix + ".rados.write_time", stats['write_time'], now)
        collect_metric(conf.graphite_host, conf.graphite_port, conf.graphite_prefix + ".rados.read_time", stats['read_time'], now)
    except Exception as e:
        raise