failed to start sisp on payloadNode
Hi,
I currenty evaluating OpenClovis as ours middleware solution.
I installed OpenClovis version 2.2 and tried to use the evaluation kit as described in the Evaluation User Guide.
I have two PCs in the same subnet.
In the target.conf file I defined one as the SCNodeI0 and the other as PayloadNode0.
Then after unzipped and extracts the PayloadNode0.tgz on the PayloadNode0 I try to start the sisp daemon and receives the following errors:
root@test1 sisp]# ./etc/init.d/sisp start
Starting sisp_amf:
At least one interface (eth0) has a low value for the "unresolved
neighbour queue length" kernel parameter. This may cause problems for
this application. See the evaluation kit release notes for more
information.
insmod: error inserting '/root/sisp/modules/clIoc.ko': -1 Invalid module format
insmod: error inserting '/root/sisp/modules/clIocUdpKernelXport.ko': -1 Invalid module format
mknod: wrong number of arguments
Try `mknod --help' for more information.
mknod: wrong number of arguments
Try `mknod --help' for more information.
[ OK ]
On the SCNode0 I didnt have any problem
Maybe different kernels?
Our kernel module that handles internode communication is not loading. What version of Linux are you running (do a "uname -a")? Please check on both machines. I'm guessing that they are different.
Thanks for the fast reply
you are right the SCNodeI0 use: 2.6.11-1.1369_FC4smp
and the PayloadNode0 use: 2.6.11-1.1369_FC4
I miss it
But How can I still evaluate OpenClovis on those machines ?
Thanks
What you are asking for is called cross-compilation, although a very trivial case. Essentially, "from envrionment A, compile code for environment B". The Clovis software allows you cross-compile for a variety of different targets, even across processors! But it takes some configuration...
The eval kit was made to be simple and so some assumptions had to be made. I don't know offhand whether you can do this easily in the eval, but I will ask around. Here are some things that you might want to try:
1. Boot your multiprocessor computer in single-processor mode (SMP means > 1 processor). Many linux distros give you a choice of kernels at boot up and often one of those choices is the non-smp version of your current kernel. If you run a non-smp kernel on an smp machine it will work, the other processor will just be idle. If the non-smp kernel does not appear, you may be able to install it through your distro's package mgmt software.
2. Install Clovis and the eval kit on both machines. Don't copy .tgz files across, instead use the one built locally. I've never tried this, so no guarantees! :-)
You can also just run the eval on one machine & still get a great overview of the product.
Andrew
I am also continuing to have this "Invalid module format" problem, and I am using what I believe should be a valid combination. I am trying to build a distributed node demo of OpenClovis using the eval. I have built and am running sisp on FC6 no problem (management node). I then "make clean", and ./configure using "--with-cross-build=ia32-linux-rhel4-2.6.9 --with-kernel-build=linux-2.6.9-5.EL-smp-i686" and then rebuild and make images. I deploy images to Intel MPCBL0001 ATCA blades running CentOS4.5 (2.6.9-55.ELsmp). When I try to fire up sisp, I get module format error. Seems like struct_module version would not change from 2.6.9-5 to 2.6.9-55, but I'm not sure. Am I enabling cross-build correctly?
I would really like to get this multi-node demo working without having to build the target on the target.