<<
If the TCP part is entirely non-working, it might be better
to disable it for now rather than have it crash the machine.
So far, I have never gotten it to function correctly and it
crashes some machines almost immediately.
My vote is to disable TCP support (return not supported) - at least for now -
and possibly remove this support. Even when it is running the performance we
see precludes its use.
Johann George wrote:
That's a remote invalid request error. Were you testing
with RDMA or without?
We were using the version that runs over IB.
What user application were you using for testing?
qperf. Unfortunately the version that is included in OFED
1.3 RC2 is old due to a problem that Vlad just discovered
yesterday. We hope to have that fixed shortly and the
updated version will be part of the nightly builds. In the
meantime, you can download it from the repository and build
it:
git clone git://git.openfabrics.org/~johann/qperf
cd qperf
./autogen.sh
./configure
make
To run it, on one machine (the server), run it with no
arguments. On the other machine, run:
qperf server_nodename rds_bw
where server_nodename is replaced by the hostname of the
server. This test gives bandwidth statistics for RDS. It
worked well with RC1 but fails with RC2.
Yes, the TCP part of RDS isn't being looked after very much,
unfortunately.
If the TCP part is entirely non-working, it might be better
to disable it for now rather than have it crash the machine.
So far, I have never gotten it to function correctly and it
crashes some machines almost immediately.
If it is helpful, we may be able to set up a VPN so you can
log onto the machines and experience the problem with the
RDMA version and the crash with TCP.
Johann
_______________________________________________
ewg mailing list
ewg@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
_______________________________________________
ewg mailing list
ewg@lists.openfabrics.org
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg
