Custom Metrics

SPM lets you extend standard reports.  SPM exposes APIs and provides libraries that lets you send custom metrics (any numerical data) into SPM and graph it along other reports.

Terminology

• A "metric" is a distinct triple of (metric name, filter1 name, filter2 name).  Some examples:
  • ("cpuIdle", server="server1", null) vs. ("cpuIdle", null, server="server1") – these are 2 distinct metrics
  • ("cpuSteal", server="server1", null) vs. ("cpuIdle", server="server1", null) – these are 2 distinct metrics
  • ("diskRead", server="server1", device="/dev/xxx") vs. ("diskRead", server="server1", device="/dev/yyy") – this is the same 1 metric with two different values for filter2 named "device"
  • ("diskWrite", server="server1", device="/dev/xxx") vs. ("diskWrite", server="server1", device="/dev/yyy") – this is the same 1 metric with two different values for filter2 named "device"
  • ("diskRead", server="server1", device="/dev/xxx") vs. ("diskWrite", server="server1", device="/dev/xxx") – these are 2 distinct metrics

• A "data point" is the value associated with some (metric, filter1, filter2) combination.  Some examples:
  • ("cpuIdle", server="server1", null) = 90
  • ("cpuIdle", server="server1", null) = 90 – this is a distinct data point for the same metric as in the previous line

Introduction

A data point has the following attributes: 

• timestamp - time since "Epoch" in milliseconds
• name - name of metric
• value - double value
• aggregation type - min, max, avg, sum
• filter1 value - string value for filter1
• filter2 value - string value for filter2

Custom metrics reports display values aggregated using 'aggregation type'. For example, values 1.0, 2.0, and 3.0 sent during the same minute with average ('avg') aggregation type will be represented by a single point on a graph, with value 2.0 for that minute.

Filter1 and filter2 values can be used to organize complex metrics.  For example, one can use them to track the number of registered users by gender and account type. In registered users example data points can be specified as: 

{
   "name" : "registered-users-count", 
   "value" : 42, 
   "aggregation" : "sum",
   "filter1" : "user.gender=male",
   "filter2" : "account.type=free"
}
 
{
   "name" : "registered-users-count", 
   "value" : 24, 
   "aggregation" : "sum",
   "filter1" : "user.gender=female",
   "filter2" : "account.type=free"
}
 
{
   "name" : "registered-users-count", 
   "value" : 10, 
   "aggregation" : "sum",
   "filter1" : "user.gender=female",
   "filter2" : "account.type=paid"
}

In reports, the selected values in one filter are combined with an 'OR', while selected values in different values are 'ANDed'. For example, to display the number of paying female users, in filter1 the value 'female' should be selected, and in filter2 value 'paid' should be selected.

We recomend to use 'parameter.name=value' format to define filter values.

Custom Metrics API

SPM provides REST API for sending data points for custom metrics:

http://spm-receiver.sematext.com/receiver/custom/receive.[format]?token=[spm app token]

The format can be either 'json' or 'raw'.

Applications' tokens are listed at https://apps.sematext.com/users-web/home.do?homeTab=spm

JSON format

URL: http://spm-receiver.sematext.com/receiver/custom/receive.json?token=[spm app token]

Content-type: application/json

Method: POST

JSON body:

Data point:

Example:

{
   "datapoints" : [
      {
         "timestamp" : 1369669889927,
         "name" : "registered-users-count",
         "value" : 1,
         "aggregation" : "sum",
         "filter1" : "user.gender=male",
         "filter2" : "account.type=free"
      },
      {
         "name" : "registered-users-count",
         "value" : 1,
         "aggregation" : "sum",
         "filter1" : "user.gender=female",
         "filter2" : "account.type=paid"
      }
 ]
}

To send metrics using curl, save example above to a file named datapoints.json and submit it with the following call, using your own token:

curl -v -H 'Content-type: application/json' -d @datapoints.json http://spm-receiver.sematext.com/receiver/custom/receive.json?token=[spm app token]

One-liner example:

curl -H 'Content-type: application/json' -d '{"datapoints" : [{"name": "registered-users", "value" : 1, "aggregation" : "sum"}]}' http://spm-receiver.sematext.com/receiver/custom/receive.json?token=[spm app token]

Raw format

URL: http://spm-receiver.sematext.com/receiver/custom/receive.raw?token=[spm app token]

Content-type: text/plain

Method: POST

Each line is tab delimited sequence of fields:

(timestamp)\t(name)\t(value)\t(aggregation type)\t(filter1)\t(filter2)

Lines are separated by new line character ('\n'). Limit is 100 lines per request.

Example:

1369671381221	registered-users-count	1	sum	user.gender=male	account.type=free
1369671381221	registered-users-count	1	sum	user.gender=female	account.type=paid

To post these data using curl, save the example above (ensure tabs are preserved) to a file named datapoints.raw and use the following call with your own token:

curl -v -H 'Content-type: text/plain' -d @datapoints.raw http://spm-receiver.sematext.com/receiver/custom/receive.raw?token=[spm app token]

One-liner example:

curl -H 'Content-type: text/plain' -d "`echo -e "\tregistered-users\t1\tsum"`" http://spm-receiver.sematext.com/receiver/custom/receive.raw?token=[spm app token]

Java API

Sematext-metrics is an open source Java library for sending custom metrics from Java applications. Please refer to README for a quick start. Sematext-metrics-reporter is Sematext-metrics reporter for Codahale Metrics Library.

Limitations

Each SPM plan allows for a different number of metrics and a different number of datapoints per calendar month.  Please see SPM Plans page.