[henrik]

Dear Tomislav,

it's always interesting to see what people use your developments for ... We never thought of this applications, but there is no reason why it should not work with some additional modules, care and attention.

@nfs: The nfs module is shipped with the distro and I used it create the distro. Simply mount directories as usual (and with the usual problem).

@sshd: It's shiped with SLAX, it not switched on by default. See here:

http://www.slax.org/modules.php?search=sshd&category=

@orte error: I see the same problem and have no idea right now.

I would also consider Alberto's OpenSUSE image. It's big, but probably more complete in tems of networking.

http://www.cfd-online.com/Forums/ope...se-studio.html

Henrik

[tomislav_maric]

Originally Posted by henrik

@nfs: The nfs module is shipped with the distro and I used it create the distro. Simply mount directories as usual (and with the usual problem).


@sshd: It's shiped with SLAX, it not switched on by default. See here:

http://www.slax.org/modules.php?search=sshd&category=

@orte error: I see the same problem and have no idea right now.

I would also consider Alberto's OpenSUSE image. It's big, but probably more complete in tems of networking.

http://www.cfd-online.com/Forums/ope...se-studio.html

Henrik

@nfs
I'm a reall noob in networking. why do I need to export the directories (and which directories) with nfs? The only thing I can think of is that processes on the slave node are able to write to the master node. Is this true? I thought that mpirun (TCP/IP) takes care of the write() function.

@sshd
solved, thank you for the advice. I wrote the wrong address for one host by accident.

@orte error:
The problem here is that mpirun uses sshd which in turn calls for a non interactive login bash shell. From what I've been reading in bash man pages, bash uses different configuration files, depending on the manner it is called (/etc/profile for all, for interactive login or non interactive with --login option, ~/.bash_profile then ~/.bash_login, ... and so on).

what I did, in my newbish way is this: I've copied the commands that source /OpenFOAM/OpenFOAM-1.5-dev/etc/bashrc and all else from .bash_profile to EVERY config bash script I could find. It didn't help at all.

First of all, I can't find where the interactive non-login bash is started, on the main node, or on the slave? And is there some recognition condition for when the bash is in non interactive mode? Then I could tell it to source /OpenFOAM/OpenFOAM-1.5-dev/etc/bashrc script and set everything right.

I've tried using mpirun this way

mpirun --prefix=/path/to/Open MPI/on/SLAX/Live -H mario -np 2 interFoam -parallel

or

/pathname/of/mpirun -H hostname -H mario -np 2 interFoam -parallel

and then I got an error that mpirun daemon terminated because he couldn't find interFoam executable on host marija (master node), which was weird.

Then I tried passing enviromental variables to the slave node with -x command and using the instructions for mpirun calls found on OpenCFD site:

/pathname/of/mpirun -x LD_LIBRARY_PATH=$LD_LIBRARY_PATH, -x PATH=$PATH -H mario -np 2 `which interFoam` -parallel

and this gave me this kind of error message:

mpi has noticed that a process on host mario has terminated with escape signal 11 (Segmentation Fault)

If I'm not hit in the head too much, after 15 rounds in the ring with mpirun, this could mean that the process started on mario slave node, and terminated for some other reasons. Perhaps I should use -x and pass other FOAM variables such as $FOAM_MPI_LIBBIN and similar.

@suseLIVE
I'm looking into it right now.

My problem is that I'm really new to networking so it's hard for me to know which way to go when an error occurs. If it were C++, there would be NO problem. I believe that everything is ok now in the networking part of the woods, the problem is with non-interactive login shell, at least that's what I've read.

thank You VERY much for the advice!

[tomislav_maric]

oh, I forgot: I've tried setting

distributed yes;

in decomposePar, so that the writing takes place on the RAMDISK on both nodes, just to see if it would work.

of course, it didn't work. what a suprise.

[andersking]

Hi,

I use the following in /etc/ssh/sshd_config to pass the required FOAM environment variables. Placed after "AcceptEnv LANG LC_*" (at least on ubuntu).

AcceptEnv PATH
AcceptEnv LD_LIBRARY_PATH
AcceptEnv FOAM_*
AcceptEnv WM_*
AcceptEnv OPAL_PREFIX
AcceptEnv MPI_BUFFER_SIZE

also when starting I use

mpirun -np 2 -host host1,host2 $(which icoFoam) -parallel

I also set up passwordless ssh login between the nodes.

Not sure exactly where sshd_config lives in SLAX though. but the above should work.

Cheers,
Andrew

[tomislav_maric]

Originally Posted by andersking

Hi,
I use the following in /etc/ssh/sshd_config to pass the required FOAM environment variables.

I have executed printenv > ~/.ssh/environment and copied it to the other host. Also I've set in /etc/ssh/sshd_config "PermitUserEnvironment" to "yes".

It works now, but I still have some questions regardint mpirun.

Question 1:
If I only have damBreak directory on my master node, then OpenFOAM starts the parallel simulation, but exits with an error message telling me it cannot read decomposeParDict. Why?

Question 2:
When I scp damBreak on the slave node from the master node and execute my command explained below, data only gets written on the slave node. Why?

What happens if I want to run the simulation on 12 slave nodes at the computer lab, will the data be written on all of them and not on the master node?



My command is:

$MPI_ARCH_PATH/bin/mpirun -np 2 -H mario `which interFoam` -parallel

Thank you very much Andrew for your advice!

Best regards,
Tomislav

[andersking]

Hmmm, not sure about this one. We generally use an nfs mounted partition for the parallel runs.

From the command that you have run - I'm assuming that all your data ended up on 'mario', and mario isn't the master node.

In this situation (**note that I haven't tested this**). You should be able to copy the processor{1,2,...,n} directories from each of the slave nodes to the master node and then use reconstructPar to put them back together.

Unfortunately you are at the limit of my parallel experiences. So there's a bit of guesswork here.

Cheers
Andrew

[tomislav_maric]

Originally Posted by andersking

From the command that you have run - I'm assuming that all your data ended up on 'mario', and mario isn't the master node.

Yes, 'mario' is the slave.

Originally Posted by andersking

In this situation (**note that I haven't tested this**). You should be able to copy the processor{1,2,...,n} directories from each of the slave nodes to the master node and then use reconstructPar to put them back together.

I was thinking about the same thing, only - it would be less time consuming if all the data is stored on the master, since the cases will be quite big.

Thank you very much for your advice!

Tomislav

[astamanea]

Originally Posted by henrik

Dear All,

there are issues with deleting and updating large files on Sourceforge. A beta-version of SLAX Distro enhanced with OpenFOAM modules which we prepared for the upcoming OpenFOAM Workshop in Montreal can be found here:

http://powerlab.fsb.hr/ped/kturbo/OpenFOAM/release/OpenFOAM_SLAX.tgz (822MB)
http://powerlab.fsb.hr/ped/kturbo/OpenFOAM/release/OpenFOAM_SLAX_noDebug.tgz (562MB)

The distribution boots from USB/DVD/CD and may be extended by adding modules from www.slax.org or other sources.

Have fun and let us know if works (or not).

Henrik

--- Instructions for a bootable USB-Stick ---

cd /media/disk
tar xfz OpenFOAM_SLAX.tgz
cd boot
./bootinst.sh (bootinst.bat in Windows)

--- Instructions for a bootable CD/DVD ---

Make an iso out of the tgz by executing

tar xfz OpenFOAM_SLAX.tgz
cd slax
make_iso.sh (slax/make_iso.bat in Windows)

and burn the result.

--- Content ---

OpenFOAM:
24M OpenFOAM-1.5-dev.General_r1238.lzm
33M OpenFOAM-1.5-dev.linuxGccDPOpt_r1238.lzm
262M OpenFOAM-1.5-dev.linuxGccDPDebug_r1238.lzm (only full release)
380K PyFoam-0.5.2.lzm
83M ThirdParty.General_2009-05-09.lzm
68M ThirdParty.linux_2009-05-09.lzm

Tools:
15M blender-2.47.lzm
6.8M engrid.lzm
6.2M ghostscript-8.61.lzm
588K gnuplot-4.2.4.lzm
1.5M grace-5.1.21-i486.lzm
904K gsl-1.8.lzm
6.4M imagemagick-6.4.8-3.lzm
4.6M mplayer-1.0.lzm
42M tetex-3.0-5.lzm

Editors:
19M emacs-cvs-23.lzm
1.3M geany.lzm
96K leafpad-0.8.16-i486-1as.lzm
6.8M vim-7.1.lzm

and many more.

Hello to everybody,
I've tried to make a bootable USB pen, but during unpacking, I get this message: tar: : .......... Cannot change ownership to uid 1000, gid 100: Operation not permitted"
could anybody help me?

thanks in advance

[fuzzysphere]

Greetings. I am trying to use the SLAX version of OPENFoam, but I am having some difficulties.

After downloading, I burned the CD as instructed, and found 3 directories on the CD: boot, slax, and an empty conference directory.

However, I can't find any further instructions as to how proceed. I looked through the directories, but cannot find anything.

Any help would be greatly appreciated! Thank you.

Sergio P.

[henrik]

Sergio,

I understand that you made an ISO out of the directories that you found in the tar and burned that. Is this correct?

If NO: please read #3 more carefully.

If YES: What happens when you boot the CD?

Henrik

[fuzzysphere]

Thanks Henrik,

Apparently God did not encumber me with excess brains - I was not booting up properly.

All apears OK now.

Thanks for the quick response.

Sergio Perez

[fuzzysphere]

Last night I tried the first OpenFOAM tutorial using SLAX, and it worked great.

Thanks for creating the SLAX version - it's been a huge help in getting OpenFOAM running!

A very basic tip to those who, like myself, are not the most computer literate: to boot up the disk, your computer may not automatically do so when you start the computer with the disk in it. You may need to press F2 repeatedly after starting the computer to get into BIOS. A screen will appear, and go to "Boot Sequence" (or something like it). Make sure that CD/DVD/CD is first on the list in the boot sequence. Then save it as you exit from BIOS.

Also, in the first OpenFOAM tutorial, the procedure from the OpenFOAM website calls for changing the directory near the beginning to something/incompressible/icoFoam... it appears there is no incompressible directory, but leaving out the incompressible part appears to work fine.

Again, many grateful thanks to the creators of OpenFOAM and the SLAX version.

Sergio Perez

[Hasselhoff]

Hi there,
first off I really appreciate this OF SLAX Version, it is a great chance for OF to attract new users due to it's simplicity I guess.

How do you deal with the data storage? Is it possible to shift the time-step folders to an internal disc or another usb, or do I have to store all data on the usb I'm working from?

Thanks in advance and best regards,
Michael

[henrik]

Dear Michael,

thanks for your interest in OpenFOAM on SLAX.

This release is about a year old. Please have a look here

http://www.cfd-online.com/Forums/ope...d-kubuntu.html

I think this release will all your requirements.

Best Regards,

Henrik Rusche

[Hasselhoff]

Thanks Henrik. I am now working with this version.
best regards,
Michael