System.Remote.Monitoring.Carbon

This module lets you periodically flush metrics to a Graphite Carbon backend. Example usage:

main = do
  store <- newStore
  forkCarbon defaultCarbonOptions store

You probably want to include some of the predefined metrics defined in the ekg-core package, by calling e.g. the registerGcMetrics function defined in that package.

Documentation

data CarbonOptions

Options to control how to connect to the Carbon server and how often to flush metrics. The flush interval should match the shortest retention rate of the matching retention periods, or you risk over-riding previous samples.

Constructors

CarbonOptions

Fields

• host :: !Text

  The hostname or IP address of the server running Carbon.

• port :: !Int

  Server port of the TCP line receiver interface.

• flushInterval :: !Int

  The amount of time between sampling EKG metrics and pushing to Carbon.

• prefix :: !Text

  Prefix to add to all metric names.

• suffix :: !Text

  Suffix to add to all metric names. This is particularly useful for sending per host stats by settings this value to: takeWhile (/= '.') <$> getHostName, using getHostName from the Network.BSD module in the network package.

Instances

Eq CarbonOptions

Methods

(==) :: CarbonOptions -> CarbonOptions -> Bool

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

Show CarbonOptions

Methods

showsPrec :: Int -> CarbonOptions -> ShowS

show :: CarbonOptions -> String

showList :: [CarbonOptions] -> ShowS

defaultCarbonOptions :: CarbonOptions

Defaults:

• host = "127.0.0.1"
• port = 2003
• flushInterval = 1000
• Empty prefix and suffix.

forkCarbon :: CarbonOptions -> Store -> IO ThreadId

Create a thread that periodically flushes the metrics in Store to Carbon. If the thread flushing statistics throws an exception (for example, the network connection is lost), this exception will be thrown up to the thread that called forkCarbon. For more control, see forkCarbonRestart.

forkCarbonRestart :: CarbonOptions -> Store -> (SomeException -> IO () -> IO ()) -> IO ThreadId

Create a thread that periodically flushes the metrics in Store to Carbon. If the thread flushing statistics throws an exception (for example, the network connection is lost), the callback function will be invoked with the exception that was thrown, and an IO computation to restart the handler.

For example, you can use forkCarbonRestart to log failures and restart logging:

forkCarbonRestart defaultCarbonOptions
                  store
                  (\ex restart -> do hPutStrLn stderr ("ekg-carbon: " ++ show ex)
                                     restart)