hriemann

A Riemann Client for Haskell

Monthly Downloads: 19

Programming language: Haskell

License: MIT License

Tags: Web     Monitoring    

Latest version: v0.3.3.2

README

A Riemann Client for Haskell

This is a Riemann client for Haskell with an API based on my experience using Riemann in a production environment.

It was initially inspired by riemann-hs however I felt things could be a little simpler.

• No Lenses
• Doesn't use a monad transformer
• Currently this library is only for sending events, not querying
• No UDP client
• Async TCP sending
• Batching and async error handling
• Deal with back pressure

Please be aware that this is currently a work in progress and hasn't been well tested yet. I still have the following features to add:

• Tests!!!
• Timeout if events take too long to send

Use pipes?

Originally I wanted to avoid pipes as it's just another dependency and Unagi seems to be quite a bit faster, however after implementing most of the required features I can see that it would be really nice if the 'Clients' were made of individual components that compose. You would have 1 producer which is connected to sendEvent, consumers such as a TCP consumer and a stdout consumer as well as pipes such as batch pipe and overflow pipe.

Usage

module Main where

import           Data.Function
import qualified Network.Monitoring.Riemann.Event as Event
import           Network.Monitoring.Riemann.TCP
import qualified Data.Sequence as Seq

main :: IO ()
main = do
    c <- tcpConnection "localhost" 5555
    putStrLn "doing some IO work"
    let event1 = Event.ok "my service"
                   & Event.description "my description"
                   & Event.metric (length [ "some data" ])
                   & Event.ttl 20
                   & Event.tags [ "tag1", "tag2" ]
    Event.sendEvents c (Seq.singleton event1)
    putStrLn "do something else"
    let event2 = Event.ok "my other service"
    Event.sendEvents c (Seq.singleton event2)
    putStrLn "finished"

Events are built by composing helper functions that set Riemann fields and applying to one of the Event constructors:

  Event.ok "my service"
& Event.description "my description" 
& Event.metric (length [ "some data" ]) 
& Event.ttl 20 
& Event.tags ["tag1", "tag2"] 

With this design you are encouraged to create an event with one of Event.ok, Event.warn or Event.failure.

This has been done because we found that it is best to avoid services like my.service.success and my.service.error (that's what the Riemann state field is for).

You can use your own states using Event.info & Event.state "trace" however this is discouraged as it doesn't show up nicely in riemann-dash.

Alternatively there is a Monoid based api for creating events:

import Data.Monoid (Endo, (<>))
import Network.Monitoring.Riemann.Event.Monoid


eventA :: Endo Event
eventA = ttl 10 <> metric (1 :: Int) <> attributes [attribute "something" Nothing]

eventB :: Endo Event
eventB = tags ["tag 1"]

compositeEvent :: Event
compositeEvent = ok "some service" (eventA <> eventB)