Re: CAO remoting and Web gardens

From: Ken Kolda (ken.kolda_at_elliemae-nospamplease.com)
Date: 09/24/04


Date: Fri, 24 Sep 2004 12:45:41 -0700

If your client can access the web servers directly (as opposed to only
through a virtual IP that maps through a load balancer), the having a web
farm should work as you noted. A web garden is different though because
there are two processes that effectively share a single listener (IIS). Now
IIS doesn't know anything about remoting -- it just knows when it receives a
request to forward it to the to pass the request to the two processes in
round robin fashion -- it's not inspecting the requests' contents to see if
it's a remoting request and, if so, which process it should go to.

Ken

<SolDel@mercer.com> wrote in message
news:Ow5RsLmoEHA.1800@TK2MSFTNGP15.phx.gbl...
> Thanks Ken.
>
> I fully understand the concept of web gardens as relates to worker
> processes. One thing that I failed to mention was that this works on four
> servers in an NLB cluster. The fifth server is where we are having an
> issue. I find it hard to believe that the first four servers work as a
mere
> coincidence - but I can find no other reason.
>
> My observations indicate that when a CAO is marshaled, it uses the primary
> IP address on the primary adapter. In an NLB environment this is
typically
> the machines individual public IP address - which provides the required
> server affinity. My assumption was that there *may* be a similar
mechanism
> in the remoting infrastructure to supply process affinity so that
subsequent
> request for the same marshaled URI are routed to the same process (instead
> of following the round robin distribution).
>
> At this point, I have four machines that work, and two that do not. I am
> looking for some empirical evidence that can explain the discrepancy. Am
I
> just lucky in the case of the four? I bet not.
>
> Dolph
>
> "Ken Kolda" <ken.kolda@elliemae-nospamplease.com> wrote in message
> news:OTHPCwkoEHA.3900@TK2MSFTNGP10.phx.gbl...
> > If you're using CAOs from within an IIS-based remoting app then you
> > shouldn't use web gardens. When you use a web garden you are instructing
> > IIS
> > to spin up several distinct processes to handle your requests. Incoming
> > requests are then sent to these processes in a round robin fashion. Your
> > CAO
> > will live in whatever process it was first instantiated in. If
subsequent
> > requests come in for your CAO and those requests are sent to the wrong
> > process, the CAO can't be found so you get the "Requested service not
> > found"
> > error.
> >
> > Ken
> >
> >
> > <SolDel@mercer.com> wrote in message
> > news:uRTS$MkoEHA.2588@TK2MSFTNGP12.phx.gbl...
> >> I have come across an issue when using CAOs and enabling the Web garden
> >> on
> >> the process model. If gardening is disabled, eveything works fine. If
> >> it
> >> is enabled, the remote object is created and appears to marshal just
> >> fine,
> >> however, any calls on the proxy result in the standard error,
> >>
> >> System.Runtime.Remoting.RemotingException: Requested Service not
> > found.
> >>
> >> First, are CAOs supported under the Web garden process model? Second,
if
> >> supported, are there any special configurations that are required? Or
> >> any
> >> known issues?
> >>
> >> I get the same results on both of the following configurations:
> >> Windows 2000 w/ .NET 1.1
> >> Windows XP w/ .NET 1.1 SP1
> >>
> >> Thanks,
> >> Dolph Priest
> >>
> >>
> >
> >
>
>



Relevant Pages

  • Re: CAO remoting and Web gardens
    ... I fully understand the concept of web gardens as relates to worker ... I find it hard to believe that the first four servers work as a mere ... My observations indicate that when a CAO is marshaled, ... > requests are then sent to these processes in a round robin fashion. ...
    (microsoft.public.dotnet.framework.remoting)
  • Re: apache httpd performance
    ... Both servers share same dns server, so I do not think that the DNS can be issue. ... We'd be better off ignoring the Linux results entirely and simply try instead to optimize your FreeBSD performance. ... My big surprise was that CPU: AMD Athlon64 Processor 3200+ can't handle 100 concurrent requests normally e.g. the load on the servers goes very high, and the server does not respond very well during the test. ...
    (freebsd-performance)
  • Re: Using external IP/DNS name for accessing internal resources
    ... Is 100.100.100.100 an IP owned by ISA (used for publishing ... When the proxy relays the requests to other servers, ...
    (microsoft.public.isa.publishing)
  • Re: What is this?
    ... >This event is generated when TCP traffic to port 0 is detected. ... This fails on a properly set up firewall. ... accessible DNS servers - one in the DMZ, and two located at our upstream. ... All internal DNS requests go to servers behind the firewall, ...
    (comp.security.firewalls)
  • Re: Memory Manager Tests - using 512 Threads
    ... > In this test, Clients are virtualy flooding the Server with requests, ... > Thread pool in RTC is growing on-demand, ... > long-running requests won't block the rest of the communication. ... > I've set up the Default value to 256, which is what most Web Servers ...
    (borland.public.delphi.thirdpartytools.general)

Quantcast