Class: Statsy::Client

Inherits:
Object
  • Object
show all
Defined in:
lib/statsy.rb

Instance Method Summary collapse

Constructor Details

#initialize(transport = Transport::UDP.new("stats", 8125)) ⇒ Client

Construct a client with a given transport that implements Transport::Interface

Usage:

client = Statsy::Client.new
client = Statsy::Client.new(Statsy::Transport::UDP.new("custom", 8888))



48
49
50
 
# File 'lib/statsy.rb', line 48

def initialize(transport=Transport::UDP.new("stats", 8125))
  @transport = transport
end

Instance Method Details

#batch {|self.class.new(batch = Transport::Queue.new)| ... } ⇒ Object

Batch multiple transport operations, that will group any counts together and send the fewest number of packets with the counts/timers optimized at the end of the batch block.

Note: this does not attempt to fit the packet size within the MTU.

Usage:

client.batch do |batch|
  batch.increment("foo.bar", 10)
  batch.measure("bat.baz", 101)
  batch.measure("foo.bar", 101)
end

=> write "foo.bar:10|c:333|ms"
=> write "bat.baz:101|ms"

Yields:



123
124
125
126
127
128
129
130
131
132
133
134
135
136
 
# File 'lib/statsy.rb', line 123

def batch
  yield self.class.new(batch = Transport::Queue.new)

  batch.inject(Hash.new { |h,k| h[k]=[] }) do |stats, stat|
    # [ "foo.bar:10|c", "foo.bar:101|ms" ]
    key, value = stat.split(':', 2)
    stats[key] << value
    stats
  end.sort.each do |pairs|
    # [ "foo.bar", [ "10|c", "101|ms" ] ]
    @transport.write(pairs.flatten.join(":"))
  end
  self
end

#increment(stat, count = 1, sampling = 1) ⇒ Object

Increment a count optionally at a random sample rate. These keys will live under the stats and stats_counts keys.

Usage:

client.increment("coffee.single-espresso")
client.increment("coffee.single-espresso", 1)
client.increment("coffee.single-espresso", 1, 0.5) # 50% of the time



60
61
62
63
64
65
66
67
68
69
 
# File 'lib/statsy.rb', line 60

def increment(stat, count=1, sampling=1)
  if sampling < 1
    if Kernel.rand < sampling
      write(stat, count, 'c', sampling)
    end
  else
    write(stat, count, 'c', 1)
  end
  self
end

#measure(stat, time) ⇒ Object

Sample a timing. The units of the timing are up to you and your consumers, milliseconds is common.

Including the units in the key name will help communicate the units to consumers of these measurements.

The statistics will be aggregated over the sampling period configured in your statsd. By default this is every 10 seconds.

In graphite, these reports will end up under the stats.timings key.

Usage:

client.measure("foo.backendtime", response.headers["X-Runtime"].to_i)

Produces the statistics found per sampling interval.

stats.timings.foo.backendtime.count
stats.timings.foo.backendtime.lower
stats.timings.foo.backendtime.mean_90
stats.timings.foo.backendtime.upper
stats.timings.foo.backendtime.upper_90



92
93
94
95
 
# File 'lib/statsy.rb', line 92

def measure(stat, time)
  write(stat, time, 'ms', 1)
  self
end

#record(stat, gauge) ⇒ Object

Record an arbitrary value

Usage:

client.record("foo.arbitrary", 42)



102
103
104
105
 
# File 'lib/statsy.rb', line 102

def record(stat, gauge)
  write(stat, gauge, 'g', 1)
  self
end