@telemetry-js/collector-counter v0.0.6
collector-counter
Collect metrics from a counter incremented by you.
Atelemetry
plugin.
Table of Contents
Usage
Has three variants: delta
, persistent
and rate
.
const telemetry = require('@telemetry-js/telemetry')()
const counter = require('@telemetry-js/collector-counter')
const errors = counter.delta('myapp.errors.delta')
telemetry.task()
.collect(errors)
.publish(..)
// Elsewhere in your app
errors.increment(1)
The .delta
function creates a metric of which the value resets to 0 after it's been collected. I.e. its value is a delta between submissions. For this type of metric, the only relevant statistic is sum
.
If you don't want the value to be reset, which is useful to describe some "current state", use .persistent
:
const connections = counter.persistent('myapp.connections.count')
telemetry.task()
.collect(connections)
.publish(..)
// Elsewhere in your app
connections.increment(1)
connections.decrement(1)
Lastly, use .rate
for rate metrics. It behaves like .delta
in that the value is reset to 0 after it's been collected, but the relevant statistic is average
. In addition, .rate
will divide your value by the number of elapsed seconds. To explain how this works, let's add a schedule to our example:
const simple = require('@telemetry-js/schedule-simple')
const requests = counter.rate('myapp.requests.per.second')
telemetry.task()
.collect(requests)
.schedule(simple, { interval: '5m' })
.publish(..)
// Elsewhere in your app
requests.increment(1)
This schedule will ping our plugin every 5 minutes. Say we have incremented requests
a 1000 times within 5 minutes. Then the plugin will emit a metric with value 1000 / (5 * 60)
: 3.3
requests per second. To get reliable data, the schedule interval should be greater than 1 second (at least).
If we want to account for spikes and valleys within the 5 minute window, we can either increase the interval (which also affects how often the metric is published) or use the summarize
processor:
const summarize = require('@telemetry-js/processor-summarize')
telemetry.task()
.collect(requests)
.schedule(simple, { interval: '30s' })
.process(summarize, { window: '5m' })
.publish(..)
This will collect our metric every 30 seconds and publish it every 5 minutes with (among other things) a min
and max
of collected values.
API
plugin = counter.delta(metricName, options)
It is recommended to end the metric name with .delta
, to differentiate it from other types.
Options:
unit
: string, defaults tocount
statistic
: string, defaults tosum
- Other options are passed as-is to
metric
.
Metrics created by .delta
get a .statistic = 'sum'
property. Using this information, the publisher-appoptics
plugin will set the summarize_function
parameter for you, which tells AppOptics to sum (rather than average) values when rolling up high-resolution measurements. No such behavior exists in CloudWatch and by extension the publisher-cloudwatch
plugin.
plugin = counter.persistent(metricName, options)
It is recommended to end the metric name with the unit, e.g. connections.count
, size.bytes
.
Options:
unit
: string, defaults tocount
statistic
: string, defaults toaverage
- Other options are passed as-is to
metric
.
plugin = counter.rate(metricName, options)
It is recommended to end the metric name with the rate and/or unit, e.g. requests.per.second
for a unit of count/second
, received.kbps
for a unit of kilobytes/second
.
Options:
unit
: string, defaults tocount/second
statistic
: string, defaults toaverage
- Other options are passed as-is to
metric
.
plugin
This is a function to be passed to a Telemetry Task. Can be used by multiple tasks, sharing state:
telemetry.task().collect(plugin)
telemetry.task().collect(plugin)
plugin.increment(n)
Increment state value by n
, defaults to 1.
plugin.decrement(n)
Decrement state value by n
, defaults to 1.
Install
With npm do:
npm install @telemetry-js/collector-counter
Acknowledgements
This project is kindly sponsored by Reason Cybersecurity Ltd.
License
MIT © Vincent Weevers
4 years ago