OpenTracing (https://opentracing.io) for Haskell
License: Apache License 2.0
OpenTracing is an attempt to define a common API to instrument code for Dapper-style distributed tracing, abstracting over various implementations of this concept (such as Twitter's Zipkin, Uber's Jaeger, or even Apache HTrace.
Hi there, I'm attempting to send instrumentation out to zipkin.
{-# LANGUAGE OverloadedStrings #-}
module Main where
import Control.Monad.IO.Class
import Control.Monad
import Control.Concurrent
import OpenTracing
import qualified OpenTracing as OT
import qualified OpenTracing.Log as OT
import qualified OpenTracing.Zipkin as Z
import qualified OpenTracing.Zipkin.HttpReporter as ZR
zipkinTracing :: MonadIO m => OT.Sampler -> m (OT.Tracing Z.ZipkinContext MonadIO)
zipkinTracing sampler = do
let endpoint = ZR.Endpoint (Just "myservice") (Just $ read "localhost") Nothing (Just 9411)
env <- Z.newEnv sampler
renv <- liftIO $ ZR.newEnv ZR.V2 endpoint "localhost" 9411 OT.jsonMap
pure $ OT.Tracing (Z.zipkinTracer env) (ZR.zipkinHttpReporter renv)
main :: IO ()
main = do
tracing <- zipkinTracing $ constSampler True
forM_ [0..1000] . const $ traced' tracing (spanOpts "hello" mempty) $ \x -> putStrLn "hello" >> threadDelay 500000I started zipkin via the quickstart guide:
wget -O zipkin.jar 'https://search.maven.org/remote_content?g=io.zipkin.java&a=zipkin-server&v=LATEST&c=exec'
java -jar zipkin.jarand it is running on localhost:9411.
Any suggestions for how to continue? Thanks!
I tried 1e130d0 with low versions of GHC to find the minimum version I could compile with and that was ghc-8.6.1. I had this problem compiling with ghc-8.4.4:
$ cabal build all --enable-tests --enable-benchmarks
Resolving dependencies...
Build profile: -w ghc-8.4.4 -O1
...
Data/Vinyl/ARec/Internal.hs:159:3: error:
• Illegal polymorphic type:
forall s. Int -> SmallMutableArray s -> ST s ()
Perhaps you intended to use RankNTypes or Rank2Types
• In the definition of data constructor ‘ARecBuilder’
In the newtype declaration for ‘ARecBuilder’
|
159 | ARecBuilder ( forall s.
| ^^^^^^^^^^^^^^^^^^^^^^^...
Error: cabal: Failed to build vinyl-0.14.3 (which is required by
exe:opentracing-example-simple from opentracing-examples-0.2.0 and
exe:opentracing-example-rpc from opentracing-examples-0.2.0). See the build
log above for details.
Should we bump the base minimum constraint like this?
- , base >= 4.9 && < 4.17
+ , base >= 4.12 && < 4.17Same story for 64fd065, the commit hash for the 0.2.2 tag.
Is there a roadmap for a Hackage release?
I’m using this library in another library I’m developing that integrates polysemy’s free monad architecture with distributed tracing.
I’d love to help out in any way that I can to get this ready for release. It’s a really great project that I think a lot of people will benefit from!
Hi!
I'm a happy user of your library (thanks!) who'd like to migrate to GHC 9.4 and 9.6.
Are you planning releasing the current master to Hackage in the near future? :)
Thanks!
The stdTracer in OpenTracing.Standard is built using a StdEnv, and the envPRNG field in StdEnv is a GenIO.
GenIO isn't thread-safe unfortunately: https://hackage.haskell.org/package/mwc-random-0.14.0.0/docs/System-Random-MWC.html#t:Gen
We noticed issues in tests that concurrently start up traces using the same Tracer where the resulting finished spans would occasionally have incorrect parent/child references due to the IDs being duplicated.
SimSpace has made the GenIO access thread-safe in this commit - https://github.com/Simspace/opentracing/commit/43ba8fd415562281e087d1d7a5b300bbd9d9837a - and our tests are working great. We're happy to upstream this fix as well.
This line adds a "kind": "foo" key to the JSON blob which is sent to Zipkin:
Curling these payloads to zipkin directly shows the issue:
$ curl -X POST --header 'content-type: application/json' --url 'http://localhost:9411/api/v2/spans' --data '[{"name":"apikinesisworkflowevents","parentId":"669d318a1f17eca2","id":"669d318a1f17eca1","traceId":"8ddd20a39da03fb1a7a7ddffe46f8747", \
"kind":"server",\
"timestamp":1509566690478713,"duration":101327,"debug":false,"localEndpoint":{"serviceName":"apollo","ipv4":"127.0.0.1","port":9411},"annotations":[]}]'
Cannot decode spans
$ curl -X POST --header 'content-type: application/json' --url 'http://localhost:9411/api/v2/spans' --data '[{"name":"apikinesisworkflowevents","parentId":"669d318a1f17eca2","id":"669d318a1f17eca1","traceId":"8ddd20a39da03fb1a7a7ddffe46f8747", \
"timestamp":1509566690478713,"duration":101327,"debug":false,"localEndpoint":{"serviceName":"apollo","ipv4":"127.0.0.1","port":9411},"annotations":[]}]'
where the only difference is the presence of the server key.
Hey I'm in the middle of trying to get opentracing (using Jaeger) working with a small Servant based API I'm writing.
I'd like to use your library, but am not sure what the license is, would you mind setting one on the repository?
I'm trying to get the WAI integration working, and I'm almost there but I can't create the last bit I need, a Tracer value.
I have the functions I want the tracer to run, but due to the type variable on tracer's definition (I think), haskell gets confused about which type to use. Even when I specialize to IO (instead of using a forall m. MonadIO m), I get this type error:
src/Tracing/TracingBackend.hs:91:118-124: error: …
• Couldn't match type ‘m’ with ‘IO’
‘m’ is a rigid type variable bound by
a type expected by the context:
forall (m :: * -> *). MonadIO m => SpanOpts -> m Span
at /path/to/src/Tracing/TracingBackend.hs:91:97
Expected type: SpanOpts -> m Span
Actual type: SpanOpts -> IO Span
• In the ‘tracerStart’ field of a record
In the first argument of ‘return’, namely
‘(tracer
{tracerStart = startFn, tracerReport = jaegerAgentReporter j})’
In the expression:
return
(tracer
{tracerStart = startFn, tracerReport = jaegerAgentReporter j})
Here's what the code looks like:
class Tracing T
-- ... other code ...
traceApplication :: t -> Application -> IO Application
instance Tracing TracingBackend where
-- ... other code ...
-- TODO: check to ensure that app wide tracing is enabled
traceApplication t app = if appWideTracingEnabled then instrumentApp app else pure app
where
jaegerOptions = tracingSettings t
appWideTracingEnabled = tracingAppWide $ tracingCfg t
sampling = probSampler $ tracingSamplingRate $ tracingCfg t
propagation = jaegerPropagation
defaultReporter :: FinishedSpan -> IO ()
defaultReporter = makeLoggerReporter t
-- Create Standard ENV for Tracer to use, create standard tracer from the env
mkTracer :: IO (SpanOpts -> IO Span)
mkTracer = newStdEnv sampling >>= pure . stdTracer
-- Update Tracer with a JaegerAgentReporter (instead of log based one), produce the middleware
augmentTracerWithJaeger :: Tracer -> (SpanOpts -> IO Span) -> IO Tracer
augmentTracerWithJaeger tracer startFn = withJaegerAgent jaegerOptions (\j -> return (tracer { tracerStart=startFn
, tracerReport=jaegerAgentReporter j
}))
-- Use the created tracer to make the middleware
makeMiddleware :: Application -> Tracer -> IO Application
makeMiddleware app tracer = pure (opentracing tracer propagation (\activeSpan -> app))
-- Instrument an actual application with the agent reporter
instrumentApp :: Application -> IO Application
instrumentApp app = liftIO (logMsg t INFO "Instrumenting application...")
>> mkTracer
>>= augmentTracerWithJaeger (Tracer undefined undefined)
>>= makeMiddleware app
>>= pureTo me it seems like haskell is expecting the IO to be an m (as bound by tracer's definition), and cannot get over the requirement.
I've tried setting ScopedTypeVariables (and abstracting the MonadIO m type in the typeclass code too), but it still won't convince Haskell that both ms are the same, I just get the same error with m1 vs m instead of m and IO.
I've looked at Backends.hs and I'm honestly not sure how it works when you're in the IO monad yet manage to create a Tracer properly.
Is there any collaboration between these 2 projects?
I think at this point it’s safe to say the library is production ready. It’s great, really.
With your recommendation in #16, I started trying out using the zipkin v2 HttpReporter with the jaeger all-in-one docker container. Things seem to work fairly well but every now and then I'm getting:
Unable to process request body: validation failure list:
parentId in body should be at least 16 chars long
Heres the payload its generating:
[{"name":"range-server response ","id":"864aec327fbcb730","traceId":"18ceef753bb8676f2acbb26ea606b7f","parentId":"d89caaa049427f8","timestamp":1529711969086253,"duration":8,"debug":false,"localEndpoint":{"serviceName":"portal-server","ipv4":"127.0.0.1"},"annotations":[],"tags":{}}]The parentId looks like its 15 characters, but it appears to be validating for 16.
According to opentracing.io the OpenTracing project is now archived and OpenTelemetry is the new version of that. Is the opentracing package compatible with OpenTelemetry?
So after getting opentracing-wai working and sending data over UDP to a Jaeger agent, I found that the emitBatch commands that are being sent aren't being properly parsed by Jaeger...
Here's waht the data looks like, and the error:
*jaeger.AgentEmitBatchArgs field 0 read error: EOF
This is in the Process function of jaeger/agent.go in the jaeger codebase (I built it locally from source to do some print-debugging.
I know the this library is still in the early stages, but is this a known bug? Has UDP to Jaeger over Thrift compact protocol worked yet?
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
Django
The Web framework for perfectionists with deadlines.
Laravel
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
Microsoft
Open source projects and samples from Microsoft.
Google
Google ❤️ Open Source for everyone.
Alibaba
Alibaba Open Source for everyone
D3
Data-Driven Documents codes.
Tencent
China tencent open source team.