Also, the problem doesn't occur with RH9 and DX4.2. In addition, trying 8x
on the buffer size does not help.
Lloyd A Treinish/Watson/[EMAIL PROTECTED]@opendx.watson.ibm.com on 10/17/2003
02:19:12 PM
Please respond to [email protected]
Sent by: [EMAIL PROTECTED]
To: [email protected]
cc:
Subject: [opendx-users] socket buffer size, again
It looks like this requires some revisiting...
I've run into a situation with dx 4.3 on RH9.0, where the problem is not
solved through the usual means. On RH8 with DX4.2, the usual solution
worked with "large" networks of various sizes, as has been discussed
before. When I moved to RH9 and DX4.3, it does not always work. For one,
"very large" network, the same deadlock occurs, even when I increased the
buffersize to 4x the usual recommendation. The offending network is about
2x bigger than the next largest network that works. I can work around this
in a clumsy fashion, by starting the exec separately and connecting
manually, sometimes requiring a couple of tries to avoid the deadlock.
I'm not sure what was changed that caused the problem characteristics to be
different.
--------------------------
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [opendx-users] strange problem with VPE
To: [email protected]
Subject: Re: [opendx-users] strange problem with VPE
From: Lloyd A Treinish <[EMAIL PROTECTED]>
Date: Sun, 12 Oct 2003 19:54:40 -0400
Importance: Normal
Reply-To: [email protected]
Sender: [EMAIL PROTECTED]
It should work in the normal way, once the buffer sizes are increased and
the environment variable set. So, I'm not sure. Rebooting might be
required to ensure the larger buffer size is available to your dx
processes.
geenen <[EMAIL PROTECTED]>@opendx.watson.ibm.com on 10/07/2003 03:49:21 PM
Please respond to [email protected]
Sent by: [EMAIL PROTECTED]
To: [email protected]
cc:
Subject: Re: [opendx-users] strange problem with VPE
Your suggestion was right but I dont no why
I started the dxui by itself
/usr/local/dx/bin_linux/dxui
now I can open my .net file and run it.
I am not able to do the same through dxexec although I did change the
buffer
DX_SOCKET_BUFSIZE=262144
more /proc/sys/net/core/rmem_max
262144
more /proc/sys/net/core/wmem_max
262144
is there something else i have to set???
maybe increase the buffer even further?
I mean it works but why does it not work in the normal way???
cheers and thanks for the help
Thomas
On Tuesday 07 October 2003 19:40, Lloyd A Treinish wrote:
> The VPE will appear to hang if the buffer is overflowed between the ui
and
> exec. This can happen with a big net and when you hit the exec button.
If
> you manually kill the dxexec process and the ui comes back, then that's
> probably what's happening. Setting the buffer size as Mattijs suggests
> should resolve the problem.
>
>
>
>
> Mattijs Janssens <[EMAIL PROTECTED]>@opendx.watson.ibm.com on
> 10/06/2003 10:23:44 AM
>
> Please respond to [email protected]
>
> Sent by: [EMAIL PROTECTED]
>
>
> To: [email protected]
> cc:
> Subject: Re: [opendx-users] strange problem with VPE
>
>
>
> What I remember is that a lot of stuff from the .cfg file gets
> transferred (even when no data is transferred). On my Linux system I
> extend the network buffer size to 256Kb
>
> echo 262144 > /proc/sys/net/core/rmem_max
> echo 262144 > /proc/sys/net/core/wmem_max
>
> and setenv the DX_SOCKET_BUFSIZE variable to use this space
>
> setenv DX_SOCKET_BUFSIZE 262144
>
> Again, have no idea whether this has anything to do with your problem
> but elimination of possible problems can never harm.
>
> Mattijd
>
> geenen wrote:
> > yep
> > the window is no longer responding and I am unable to close the VPE
other
>
> than
>
> > kill it.
> > I do not actually load any data so there is no need for large amounts
of
>
> data
>
