A simple way to get your custom application metrics into Server Density for graphing and alerting is to use our custom statsd implementation.
Installing the Sdstatsd Service
Install the Sdstatsd service on Debian/Ubuntu:
sudo apt-get install sd-agent-sdstatsd
Install the Sdstatsd service on RHEL/CentOS:
sudo yum install sd-agent-sdstatsd
Configuring the agent to monitor via Sdstatsd
By default the agent Sdstatsd does not require any configuration. It will automatically bind to localhost or 127.0.0.1 on port 8125 on the next agent restart.
You can configure the following options in your sd-agent config.cfg to customise your sdstatsd deployment:
# If you don't want to enable the SdStatsd server, set this option to no
# use_sdstatsd: yes
# Make sure your client is sending to the same port.
# sdstatsd_port: 8125
# you may want all statsd metrics coming from this host to be namespaced
# in some way; if so, configure your namespace here. a metric that looks
# like `metric.name` will instead become `namespace.metric.name`
# statsd_metric_namespace:
# By default, sdstatsd supports only plain ASCII packets. However, most
# (sd)statsd client support UTF8 by encoding packets before sending them
# this option enables UTF8 decoding in case you need it.
# However, it comes with a performance overhead of ~10% in the sdstatsd
# server. This will be taken care of properly in the new gen agent core.
# utf8_decoding: false
# The number of bytes allocated to the statsd socket receive buffer. By default,
# this value is set to the value of `/proc/sys/net/core/rmem_default`. If you
# need to increase the size of this buffer but keep the OS default value the
# same, you can set sdstats's receive buffer size here. The maximum allowed
# value is the value of `/proc/sys/net/core/rmem_max`.
# statsd_so_rcvbuf: Posting metrics to Sdstatsd
It's simple to post metrics to Sdstatsd. The python example below will send the metric 'application.metric.example' with a random value between 1-100. The metric is a counter and is tagged with 'example: tag'.
import socket
from random import randint
count = randint(1, 100)
HOST = 'localhost'
PORT = 8125
MESSAGE = 'application.metric.example:{}|c|#example: tag'.format(count)
sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
sock.sendto(MESSAGE, (HOST, PORT))Metric Types and Formatting
Your metrics should be formatted in the following manner when sending them to sdstatsd
metric.name:value|type|@sample_rate|#tag1:value,tag2
- metric.name - A string with no colons, bars, or @ characters. This is the name that your metrics will appear under in your Server Density account
- value - a float or integer.
- type -
cfor counter,gfor gauge,msfor timer,hfor histogram,sfor set. - sample rate (optional) - A float between 0 and 1, inclusive. This is only available for counter, histogram and timer metrics. Defaults to 1, which means sample 100% of the time.
- tags (optional) - a comma separated list of tags. You can use colons for key:value tags. The device key is reserved and will be removed automatically.
Some examples of correctly formatted metrics are below:
# Increment the app.logins counter
app.logins:1|c
# The cookie jar is half-full, gauge
cookie.jar:0.5|g
# Sample the video length histogram half of the time
video.length:240|h|@0.5
# Track a unique visitor to the site
users.uniques:1234|s
# Increment the application logged in users counter, tagging by country of origin
app.users.logged_in:1|c|#country:gb
# Track application logged in users, tagging by country of origin and use a 50% sample rate
app.users.logged_in:1|c|@0.5|#country:gb
Comments