Hi Elmar,
could you first update to the latest version of akka-http, 10.0.3? We fixed
a few issues with the connection pool since akka 2.4.7.
Johannes
On Wednesday, February 15, 2017 at 5:12:18 PM UTC+1, Elmar Weber wrote:
>
> Hello,
>
> we are observing an issue with akka-http client side since a few days.
> With additional logging we tracked it down to the part where akka-http
> sends out the request, i.e. the client starts the request but we never see
> it hitting the target server.
>
> The client pool is set up using the following way:
>
> private val pool = scheme match {
> case HttpScheme => Http(system).cachedHostConnectionPool[HttpRequest](host,
> port, settings = settings.getOrElse(ConnectionPoolSettings(system)))
> case HttpsScheme =>
> Http(system).cachedHostConnectionPoolHttps[HttpRequest](host, port, settings
> = settings.getOrElse(ConnectionPoolSettings(system)))
> }
>
>
> Actual call goes out like this:
>
> // req: HttpRequest
> val source = Source.single(req -> req)
> val poolFlow = source.via(pool)
> poolFlow .runWith(Sink.head).flatMap {
> case (Success(response), _) => Future.successful(response)
> case (Failure(ex), _) => Future.failed(ex)
> }
>
>
> With log statements we tracked the code until the "runWith" call. After
> this there is no exception, no timeout or any other error in the log files.
> The target server is never hit with the request.
> No changes in configuration or setup that we can attribute to this, before
> the issue appeared the service hasn't been updated for around 5 days and
> was running in general without this issue since 2 months.
> Scala 2.11, akka 2.4.7, JDK 1.8.0_121, running in a docker container on
> top of Kubernetes.
> The target server is SSL secured, so it's an HTTPS connection.
>
> Other connections to SSL secured endpoints (different servers) running
> from the same application are working even when this issue occurs, however,
> the difference is that the connection pools of the working onces are hit
> several times per minute, this endpoint maybe idle for hours and then
> re-used.
> After a restart the issue is fixed and it's working for a up to 12h, then
> it's again in this state. Locally it wasn't reproducible.
>
> What we tried so far:
>
> - adding lot's of logging to get identify the line with the runWith
> - switched to pseudo random number generator
> (-Djava.security.egd=file:/dev/./urandom), had this issue a few times with
> other secure connections on the database level that resulting in blocking
> calls
> - checked stack trace dumps, nothing except expected threads in
> waiting and default akka and mongoDB pool threads, nothing that is
> different from stack trace dump on a working system versus a non working
> system (dumps are attached if helpful), this also showed us that our
> random
> number generator issue was not the source
> - right now we added a health check that constantly makes use of the
> connection pool, to see if it has something to do with the connection not
> being used
>
> Is this an issue anyone has come across or has an idea how to further
> debug the source?
>
>
> Thanks,
>
> Elmar
>
--
>>>>>>>>>> Read the docs: http://akka.io/docs/
>>>>>>>>>> Check the FAQ:
>>>>>>>>>> http://doc.akka.io/docs/akka/current/additional/faq.html
>>>>>>>>>> Search the archives: https://groups.google.com/group/akka-user
---
You received this message because you are subscribed to the Google Groups "Akka
User List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/akka-user.
For more options, visit https://groups.google.com/d/optout.
