Hello,

  Just writing to let people know the resolution of this problem...

  After much frustration and toil, we realized there was a bug in GHC's handle abstraction over sockets.

  We resolved our immediate problem by having our code deal directly with the sockets, and we filed a bug report, #2703, which has just been (partially fixed) by Simon Marlow.

thanks,
  Jeff

Simon Marlow <simonmarhaskell@gmail.com> wrote on 10/10/2008 09:23:31 AM:

> Jeff Polakow wrote:
>
> > Don Stewart <dons@galois.com> wrote on 10/09/2008 02:56:02 PM:
> >
> >  > jeff.polakow:
> >  > >    We have a server that accepts messages over a socket, spawning
> > threads to
> >  > >    process them. Processing these messages may cause other, outgoing
> >  > >    connections, to be spawned. Under sufficient load, the main
> > server loop
> >  > >    (i.e. the call to accept, followed by a forkIO), becomes
> > nonresponsive.
> >  > >
> >  > >    A smaller distilled testcase reveals that when sufficient socket
> > activity
> >  > >    is occurring, an incoming connection may not be responded to
> > until other
> >  > >    connections have been cleared out of the way, despite the fact
> > that these
> >  > >    other connections are being handled by separate threads. One
> > issue that
> >  > >    we've been trying to figure out is where this behavior arises
> > from-- the
> >  > >    GHC rts, the Network library, the underlying C libraries.
> >  > >
> >  > >    Have other GHC users doing applications with large amounts of
> >  > socket usage
> >  > >    observed similar behavior and managed to trace back where it
> > originates
> >  > >    from? Are there any particular architectural solutions that
> > people have
> >  > >    found to work well for these situations?
> >  >
> >  > Hey Jeff,
> >  >
> >  > Can you say which GHC you used, and whether you used the threaded
> >  > runtime or non-threaded runtime?
> >  >
> > Oops, forgot about that...
> >
> > We used both ghc-6.8.3 and ghc-6.10.rc1 and we used the threaded
> > runtime. We are running on a 64 bit linux machine using openSUSE 10.
>
> The scheduler doesn't have a concept of priorities, so the accepting thread
> will get the same share of the CPU as the other threads.  Another issue is
> that the accepting thread has to be woken up by the IO manager thread when
> a new connection is available, so we might have to wait for the IO manager
> thread to run too.  But I wouldn't expect to see overly long delays.  Maybe
> you could try network-alt which does its own IO multiplexing.
>
> If you have multiple cores, you might want to try fixing the thread
> affinity - e.g. put all the worker threads on one core, and the accepting
> thread on the other core.  You can do this using GHC.Conc.forkOnIO, with
> the +RTS -qm -qw options.
>
> Other than that, I'm not sure what to try right now.  We're hoping to get
> some better profiling for parallel/concurrent programs in the future, but
> it's not ready yet.
>
> Cheers,
>    Simon

---

This e-mail may contain confidential and/or privileged information. If you
are not the intended recipient (or have received this e-mail in error)
please notify the sender immediately and destroy this e-mail. Any
unauthorized copying, disclosure or distribution of the material in this
e-mail is strictly forbidden.