wavefront-sdk
This is a Ruby SDK for v2 of Wavefront's public API. It aims to be more lightweight, consistent, simple, and convenient than an auto-generated SDK.
As well as complete API coverage, wavefront-sdk
includes methods
which facilitate various common tasks, and provides non-API
features such as credential management, and writing points through a
proxy. It also has methods mimicking the behaviour of useful v1 API
calls which did not make it into v2.
Installation
$ gem install wavefront-sdk
or to build locally,
$ gem build wavefront-sdk.gemspec
wavefront-sdk
requires Ruby >= 3.0. All its dependencies are pure
Ruby, right the way down, so a compiler should never be required to
install it.
Documentation
The code is documented with YARD and automatically generated documentation is available on rubydoc.info.
Examples
First, let's list the Wavefront proxies in our account. The list()
method will return a Wavefront::Response
object. This object has
status
and response
methods. status
always yields a structure
containing result
, message
and code
fields which can be
inspected to ensure an API call was processed successfully.
response
gives you the JSON response from the API, conveniently
processed and turned into a Map
object. Map objects can be interrogated in various ways. For
instance map['items']
, map[:items]
and map.items
will all get
you to the same place.
Standard API Calls
# Define our API endpoint. (This is not a valid token!)
CREDS = { endpoint: 'metrics.wavefront.com',
token: 'c7a1ff30-0dd8-fa60-e14d-f58f91bafc0e' }
require 'wavefront-sdk/proxy'
# You can pass in a Ruby logger object, and tell the SDK to be
# verbose.
require 'logger'
log = Logger.new(STDOUT)
wf = Wavefront::User.new(CREDS, verbose: true, logger: log)
proxies = wf.list
puts proxies.class
# Wavefront::Response
# See how things went. How specific do you want to be?
puts proxies.ok?
# true
puts proxies.empty?
# false
puts proxies.status
# {:result=>"OK", :message=>"", :code=>200}
puts proxies.status.code
# 200
# Now print the proxy IDs
puts proxies.ids
# 1439acb2-ab07-4cf9-8397-2f2d758e52a0
# 87eca9df-fc47-4a24-88cf-6dd0bae245a9
# df77bd37-8f32-4e0c-b578-51eb42f22b6f
# Delete the first one.
result = wf.delete('1439acb2-ab07-4cf9-8397-2f2d758e52a0')
puts result.ok?
# true
By default (because it's the default behaviour of the API),
all API classes (except user
) will only return blocks of results
when you ask for a list of objects.
You can set an offset and a limit when you list, but setting the
limit to the magic value :all
will return all items, without you
having to deal with pagination. When you do that, offset
is repurposed as
the number of results fetched with each call to the API.
Calling a method with the limit set to :lazy
returns a lazy
enumerable. Again, offset
is the chunk size.
wf = Wavefront::Alert.new(creds.all)
# The first argument is how many object to get with each API call,
# the second gets us a lazy #Enumerable
wf.list(99, :lazy).each { |alert| puts alert.name }
# Point Rate
# Disk Error
# ...
Credentials
The SDK provides a helper class for extracting credentials from a configuration file. If you don't supply a file, defaults will be used. You can even override things with environment variables.
require 'wavefront-sdk/credentials'
c = Wavefront::Credentials.new
# Now use that to list the alerts in our account
require 'wavefront-sdk/alert'
p Wavefront::Alert.new(c.creds).list
# To get proxy configuration, use the `proxy` method. This is
# required by the Write class. You can also use c.all, which
# includes proxy and API configuration.
wf = Wavefront::Write.new(c.proxy)
wf = Wavefront::Write.new(c.all)
Queries
We can easily write queries. Let's retrieve a timeseries over the last 10 minutes, with one minute bucket granularity. We will describe the time as a Ruby object, but could also use an epoch timestamp. The SDK happily converts between the two.
require 'wavefront-sdk/query'
Wavefront::Query.new(CREDS).query(
'ts("prod.www.host.tenant.physicalmem.usage")',
:m,
(Time.now - 600)
)
Sending Metrics
The Wavefront::Write
and Wavefront::Distribution
classes lets
you send points to Wavefront in a number of ways.
Sending Points
Use Wavefront::Write
to send points. Points are described as an
array of hashes. For example:
wf = Wavefront::Write.new(Wavefront::Credentials.new.proxy)
wf.write([{ path: 'dev.test.sdk', value: 10 }])
The point hash also accepts optional source
, ts
, and tag
keys.
tag
is a hash describing point tags. For example.
wf.write({ path: 'dev.test.sdk',
value: 10,
ts: Time.now,
source: 'example',
tags: { language: 'ruby',
level: 'beginner'})
As the example shows, if you are sending a single point, you can send a naked hash, omitting the array syntax.
By default, Wavefront::Write#write
will open a connection to
Wavefront on each call, closing it after use.
If you prefer to manage the connection yourself, supply noauto:
true
in the options hash when instantiating the Write
class.
wf = Wavefront::Write.new(Wavefront::Credentials.new.proxy, noauto: true)
wf.open
wf.write(path: 'dev.test.sdk', value: 10)
wf.close
Alternatively, pass false
as the second argument to Write#write
.
(This is the legacy method, kept in for backward compatibility.)
wf = Wavefront::Write.new(Wavefront::Credentials.new.proxy)
wf.open
wf.write([{ path: 'dev.test.sdk', value: 10 }], false)
wf.close
By default, Write#write
speaks to a TCP socket on a nearby proxy,
but other methods are supported via the writer
option.
# To send points via the API
wf = Wavefront::Write.new(Wavefront::Credentials.new.creds, writer: :api)
# To send points via a local Unix socket
wf = Wavefront::Write.new({ socket: '/tmp/wf_sock'}, { writer: :socket })
# To send points over HTTP
wf = Wavefront::Write.new(Wavefront::Credentials.new.creds, writer: :http)
# Then call wf.write as before.
Write
can output verbose and debug info, and the response object
provides a summary
object.
wf = Wavefront::Write.new(Wavefront::Credentials.new.proxy, verbose: true)
wf.write([{ path: 'dev.test.sdk', value: 11, tags: { tag1: 'mytag'} }])
# SDK INFO: dev.test.sdk 11 source=box tag1="mytag"
wf = Wavefront::Write.new(Wavefront::Credentials.new.proxy, debug: true)
wf.write([{ path: 'dev.test.sdk', value: 11, tags: { tag1: 'mytag'} }])
# SDK DEBUG: Connecting to wavefront:2878.
# SDK INFO: dev.test.sdk 11 source=box tag1="mytag"
# SDK DEBUG: Closing connection to proxy.
task = wf.write([{ path: 'dev.test.sdk_1', value: 1 },
{ path: 'dev.test.sdk_2', value: 2 }])
p task.response
# {"sent"=>2, "rejected"=>0, "unsent"=>0}
puts task.ok?
# true
You can send delta metrics my prefixing your path
with a delta
symbol, or by using the Write#write_delta()
method. This is called in
exactly the same way as Write#write
, and supports all the same
options.
If you try to send huge amounts of metrics in a single go,
Wavefront::Write
will break them up into smaller API-friendly
chunks.
Sending Distributions
Use the Wavefront::Distribution
class to send distributions via a
proxy. This is an extension of Wavefront::Write
, so usage is
almost the same. All you have to do differently is specify an
interval size (m
, h
, or d
), and use a distribution as your
value
. We give you methods to help with this. For instance:
wf = Wavefront::Distribution.new(CREDS.proxy)
dist = wf.mk_distribution([7, 7, 7, 8, 8, 9, 10, 10])
p dist
# [[3, 7.0], [2, 8.0], [1, 9.0], [2, 10.0]]
p wf.write({ path: 'dev.test.dist', value: dist, interval: :m }).response
# {"sent"=>1, "rejected"=>0, "unsent"=>0}
Metric Helpers
The Wavefront::MetricHelper
class gives you simple ways to write
metrics to in-memory buffers, and flush those buffer whenever you see
fit. It aims to be a little bit like Dropwizard.
MetricHelper
gives you less control over the metrics you send. For
instance, the source and timestamp are automatically sent. You can
view the buffer at any time with the buf
attr_accessor
.
require 'wavefront-sdk/metric_helper'
wf = Wavefront::MetricHelper.new(CREDS.proxy, verbose: true)
wf.gauge('my.gauge', 1)
wf.gauge('my.gauge', 2, { tag1: 'val1' })
wf.gauge('my.gauge', 3, { tag1: 'val2' })
wf.counter('my.counter')
wf.counter('my.counter', 1, { tag1: 'val1' } )
wf.counter('my.counter')
pp wf.buf
# {:gauges=>
# [{:path=>"my.gauge", :ts=>1548633249, :value=>1},
# {:path=>"my.gauge", :ts=>1548633249, :value=>2, :tags=>{:tag1=>"val1"}},
# {:path=>"my.gauge", :ts=>1548633249, :value=>3, :tags=>{:tag1=>"val2"}}],
# :counters=>{["my.counter", nil]=>2, ["my.counter", {:tag1=>"val1"}]=>1}}
wf.flush
# SDK INFO: my.gauge 1 1548633515 source=box
# SDK INFO: my.gauge 2 1548633515 source=box tag1="val1"
# SDK INFO: my.gauge 3 1548633515 source=box tag1="val2"
# SDK INFO: ∆my.counter 2 1548633515 source=box
# SDK INFO: ∆my.counter 1 1548633515 source=box tag1="val1"
pp wf.buf
# {:gauges=>[], :counters=>[]}
Note that gauges are sent individually, timestamped at the time they are created. All counters are aggregated as you go along, and when flushed, they send their value at that moment as a single delta metric, the timestamp being the time of the flush.
You can also work with distributions. To do this, you must add
dist_port
to your options hash, giving the number of the proxy
port listening for Wavefront format distributions. Numbers can be added
to distributions individually, or in an array. You must specify the
distribution interval.
wf = Wavefront::MetricHelper.new(CREDS.proxy, { verbose: true, dist_port: 40000 })
wf.dist('my.dist', :m, 10)
wf.dist('my.dist', :m, 10)
wf.dist('my.dist', :m, [8, 8, 8, 9, 10, 10])
wf.dist('my.dist', :m, 8)
pp wf.buf
# {:gauges=>[],
# :counters=>{},
# :dists=>{["my.dist", :m, nil]=>[10, 10, 8, 8, 8, 9, 10, 10, 8]}}
wf.flush
# SDK INFO: !M 1548634226 #4 10.0 #4 8.0 #1 9.0 my.dist source=box
Contributing
Fork it, fix it, send me a PR. Please supply tests, and try to keep Rubocop happy.