Wrecker

wrecker is a library and executable for creating HTTP benchmarks. It is designed for benchmarking a series of dependent requests.

wrecker includes a wrapped version of the wreq Session API, mainly through Network.Wreq.Wrecker.

import Wrecker to write clients and Wrecker to run the them with either defaultMain or run.

See https://github.com/lorenzo/wrecker#readme for more information.

Entry Points

defaultMain :: [(String, Environment -> IO ())] -> IO ()

defaultMain is typically the main entry point for wrecker benchmarks. defaultMain will parse all command line arguments and then call run with the correct Options.

import Wrecker
import Your.Performance.Scripts (landingPage, purchase)

main :: IO ()
main = defaultMain
 [ ("loginReshare", loginReshare)
 , ("purchase"    , purchase    )
 ]

To see the options defaultMain can parse call `--help`

$ wrecker-based-app --help

wrecker - HTTP stress tester and benchmarker

Usage: wreck [--concurrency ARG] [--bin-count ARG] ([--run-count ARG] |
             [--run-timed ARG]) [--timeout-time ARG] ([--non-interactive] |
             [--interactive]) [--log-level ARG] [--log-format ARG] [--match ARG]
             [--request-name-size ARG] [--output-path ARG] [--silent]
             ([--relative-url-display] | [--absolute-url-display])
             [--record-query] [--list-test-groups]
  Welcome to wrecker

Available options:
  -h,--help                Show this help text
  --concurrency ARG        Number of threads for concurrent requests
  --bin-count ARG          Number of bins for latency histogram
  --run-count ARG          number of times to repeat
  --run-timed ARG          number of seconds to repeat
  --timeout-time ARG       How long to wait for all requests to finish
  --log-level ARG          Log to stderr events of criticality greater than the
                           LOG_LEVEL
  --log-format ARG         Log format to use
  --match ARG              Only run tests that match the glob
  --request-name-size ARG  Request name size for the terminal display
  --output-path ARG        Save a JSON file of the the statistics to given path
  --silent                 Disable all output
  --record-query           Take in consideration the query string for the report
  --list-test-groups       Shows the list of tests to run and exit

run :: Options -> [(String, Environment -> IO ())] -> IO (HashMap String AllStats)

run is the a lower level entry point, compared to defaultMain. Unlike defaultMain no command line argument parsing is performed. Instead, Options are directly passed in. defaultOptions can be used as a default argument for run.

Like defaultMain, run creates a Recorder and passes it each benchmark.

runOne :: Options -> (Environment -> IO ()) -> IO AllStats

Run a single benchmark

Wrecker State

data Environment

The Environment holds state necessary to make and record HTTP calls.

Constructors

Environment

Fields

• recorder :: Recorder

  The Recorder can be used with the record function to ... record times.

• context :: ConnectionContext

  Provided as a convience, this is a shared TLS context to reuse for better performance.

• logger :: Logger

Recorder

data Recorder

An opaque type for recording actions for profiling. To obtain a Recorder use either run, defaultMain, runOne or newStandaloneRecorder.

newtype LogicError

Constructors

LogicError String

Instances

Show LogicError

Methods

showsPrec :: Int -> LogicError -> ShowS

show :: LogicError -> String

showList :: [LogicError] -> ShowS

Exception LogicError

Methods

toException :: LogicError -> SomeException

fromException :: SomeException -> Maybe LogicError

displayException :: LogicError -> String

record :: forall a. Recorder -> String -> IO a -> IO a

record is a low level function for collecting timing information. Wrap each action of interest in a call to record.

record recorder $ threadDelay 1000000

record measures the elapsed time of the call, and catches HttpException in the case of failure. This means failures must be thrown if they are to be properly recorded.

Options

data Options

Constructors

Options

Fields

• concurrency :: Int

  The number of simulatanous connections

• binCount :: Int

  The number of bins for the histogram

• runStyle :: RunType

  runStyle determines if the wrecker runs for a specified time period or for a specified number of runs.

• timeoutTime :: Int

  How long to wait after the first benchmark for the other threads to finish

• displayMode :: DisplayMode

  This controls the command line display. It can be either Interactive of NonInteractive

• logLevel :: LogLevel

  The log severity to shown in stderr

• logFmt :: LogFormat

  The log format to use in stderr (either Json or PlainText)

• match :: String

  Set this to filter the benchmarks using a pattern

• requestNameColumnSize :: Maybe Int

  Limit the request name column to the given size

• outputFilePath :: Maybe FilePath

  Dump the results to JSON file

• silent :: Bool

  Set silent to true to disable all output.

• urlDisplay :: URLDisplay

  Whether to display short of full URLs in the report

• recordQuery :: Bool

  Set recordQuery to consider the query string as a different URL

• listTestGroups :: Bool

  Show the list of exposed test groups, if any

Instances

Eq Options

Methods

(==) :: Options -> Options -> Bool

(/=) :: Options -> Options -> Bool

Show Options

Methods

showsPrec :: Int -> Options -> ShowS

show :: Options -> String

showList :: [Options] -> ShowS

data URLDisplay

Constructors

Path

Full

Instances

Eq URLDisplay

Methods

(==) :: URLDisplay -> URLDisplay -> Bool

(/=) :: URLDisplay -> URLDisplay -> Bool

Read URLDisplay

Methods

readsPrec :: Int -> ReadS URLDisplay

readList :: ReadS [URLDisplay]

readPrec :: ReadPrec URLDisplay

readListPrec :: ReadPrec [URLDisplay]

Show URLDisplay

Methods

showsPrec :: Int -> URLDisplay -> ShowS

show :: URLDisplay -> String

showList :: [URLDisplay] -> ShowS

data RunType

There are two typical ways to invoke wrecker. RunCount will execute each a script n times on each thread. So a run count of 100 and a concurrency of 10 will run the script a total of 1000 times. Alternatively, wrecker can run for specified number of seconds with RunTimed.

Constructors

RunCount Int

RunTimed Int

Instances

Eq RunType

Methods

(==) :: RunType -> RunType -> Bool

(/=) :: RunType -> RunType -> Bool

Show RunType

Methods

showsPrec :: Int -> RunType -> ShowS

show :: RunType -> String

showList :: [RunType] -> ShowS

data DisplayMode

DisplayMode controls how results are displayed in the console. The default is NonInterative which returns the final results at the end of the program. Interactive will show partial results as the program updates.

Constructors

Interactive

NonInteractive

Instances

Eq DisplayMode

Methods

(==) :: DisplayMode -> DisplayMode -> Bool

(/=) :: DisplayMode -> DisplayMode -> Bool

Read DisplayMode

Methods

readsPrec :: Int -> ReadS DisplayMode

readList :: ReadS [DisplayMode]

readPrec :: ReadPrec DisplayMode

readListPrec :: ReadPrec [DisplayMode]

Show DisplayMode

Methods

showsPrec :: Int -> DisplayMode -> ShowS

show :: DisplayMode -> String

showList :: [DisplayMode] -> ShowS

defaultOptions :: Options

defaultOptions provides sensible default for the Options types

data AllStats

AllStats has all of the ... stats. This type stores all of the information wrecker uses to display metrics to the user.

Constructors

AllStats

Fields

• aRollup :: !ResultStatistics

  The "total" stats. This computes things like total 2xx and average time Across all requests.

• aCompleteRuns :: !ResultStatistics

  This contains statistic for actions that completed entirely successfully. Useful for knowing if a complex action is under some desired total time.

• aRuns :: !(HashMap Int ResultStatistics)

  This is an intermediate holding spot for scripts that are still executing.

• aPerUrl :: !(HashMap String ResultStatistics)

  This is the per key (URL) statistics.

Instances

Show AllStats

Methods

showsPrec :: Int -> AllStats -> ShowS

show :: AllStats -> String

showList :: [AllStats] -> ShowS

ToJSON AllStats

Methods

toJSON :: AllStats -> Value

toEncoding :: AllStats -> Encoding

toJSONList :: [AllStats] -> Value

toEncodingList :: [AllStats] -> Encoding

data ResultStatistics

This type includes statistics for all of the result values we can detect. This type is used by AllStats to compute per key (URL) statistics among other uses.

Constructors

ResultStatistics

Fields

• rs2xx :: !Statistics
• rs4xx :: !Statistics
• rs5xx :: !Statistics
• rsFailed :: !Statistics
• rsRollup :: !Statistics
• rsTotalTests :: !Int
• rsTestsFailed :: !Int

Instances

Show ResultStatistics

Methods

showsPrec :: Int -> ResultStatistics -> ShowS

show :: ResultStatistics -> String

showList :: [ResultStatistics] -> ShowS

ToJSON ResultStatistics

Methods

toJSON :: ResultStatistics -> Value

toEncoding :: ResultStatistics -> Encoding

toJSONList :: [ResultStatistics] -> Value

toEncodingList :: [ResultStatistics] -> Encoding