Instrument.Worker

Documentation

initWorkerCSV

Arguments

:: ConnectInfo

-> FilePath

Target file name

-> Int

Aggregation period / flush interval in seconds

-> AggProcessConfig

-> IO ()

A CSV backend to store aggregation results in a CSV

initWorkerCSV'

Arguments

:: FilePath

Target file name

-> AggProcessConfig

-> IO AggProcess

Create an AggProcess that dumps to CSV. Use this to compose with other AggProcesses

initWorkerGraphite

Arguments

:: ConnectInfo

Redis connection

-> Int

Aggregation period / flush interval in seconds

-> HostName

Graphite host

-> Int

Graphite port

-> AggProcessConfig

-> IO ()

Initialize a Graphite backend

initWorkerGraphite'

Arguments

:: HostName

Graphite host

-> Int

Graphite port

-> AggProcessConfig

-> IO AggProcess

Crete an AggProcess that dumps to graphite. Use this to compose with other AggProcesses

work :: Connection -> Int -> AggProcess -> IO ()

Go over all pending stats buffers in redis.

initWorker :: String -> ConnectInfo -> Int -> AggProcess -> IO ()

Generic utility for making worker backends. Will retry indefinitely with exponential backoff.

data AggProcess

A function that does something with the aggregation results. Can implement multiple backends simply using this. Note that Semigroup and Monoid instances are provided for defaulting and combining agg processes.

Constructors

AggProcess

Fields

• apConfig :: AggProcessConfig
• apProc :: Aggregated -> Redis ()

Instances

Monoid AggProcess

Defined in Instrument.Worker

Methods

mempty :: AggProcess

mappend :: AggProcess -> AggProcess -> AggProcess

mconcat :: [AggProcess] -> AggProcess

Semigroup AggProcess

Defined in Instrument.Worker

Methods

(<>) :: AggProcess -> AggProcess -> AggProcess

sconcat :: NonEmpty AggProcess -> AggProcess

stimes :: Integral b => b -> AggProcess -> AggProcess

Configuring agg processes

data AggProcessConfig

General configuration for agg processes. Defaulted with def, defAggProcessConfig, and mempty. Configurations can be combined with (<>) from Monoid or Semigroup.

Constructors

AggProcessConfig

Fields

• metricQuantiles :: MetricName -> Set Quantile

  What quantiles should we calculate for any given metric, if any? We offer some common patterns for this in quantileMap, standardQuantiles, and noQuantiles.

Instances

Monoid AggProcessConfig

Defined in Instrument.Worker

Methods

mempty :: AggProcessConfig

mappend :: AggProcessConfig -> AggProcessConfig -> AggProcessConfig

mconcat :: [AggProcessConfig] -> AggProcessConfig

Semigroup AggProcessConfig

Defined in Instrument.Worker

Methods

(<>) :: AggProcessConfig -> AggProcessConfig -> AggProcessConfig

sconcat :: NonEmpty AggProcessConfig -> AggProcessConfig

stimes :: Integral b => b -> AggProcessConfig -> AggProcessConfig

Default AggProcessConfig

Defined in Instrument.Worker

Methods

def :: AggProcessConfig

standardQuantiles :: MetricName -> Set Quantile

This is usually a good, comprehensive default. Produces quantiles 10,20,30,40,50,60,70,80,90,99. *Note:* for some backends like cloudwatch, each quantile produces an additional metric, so you should probably consider using something more limited than this.

noQuantiles :: MetricName -> Set Quantile

Regardless of metric, produce no quantiles.

quantileMap

Arguments

:: Map MetricName (Set Quantile)

-> Set Quantile

What to return on miss

-> MetricName -> Set Quantile

If you have a fixed set of metric names, this is often a convenient way to express quantiles-per-metric.

defAggProcessConfig :: AggProcessConfig

Uses standardQuantiles.

Exported for testing

expandDims :: forall packets. (Monoid packets, Eq packets) => Map Dimensions packets -> Map Dimensions packets

Take a map of packets by dimensions and *add* aggregations of the existing dims that isolate each distinct dimension/dimensionvalue pair + one more entry with an empty dimension set that aggregates the whole thing. worked example:

Given: { {d1=>d1v1,d2=>d2v1} => p1 , {d1=>d1v1,d2=>d2v2} => p2 } Produces: { {d1=>d1v1,d2=>d2v1} => p1 , {d1=>d1v1,d2=>d2v2} => p2 , {d1=>d1v1} => p1 + p2 , {d2=>d2v1} => p1 , {d2=>d2v2} => p2 , {} => p1 + p2 }