test of install on mars

Attached is a set of notes on my attempt to install from binary on mars.
Bottom line:
* kernel was easy to install and runs fine out of the box
* visit is (so far) impossible
* remote installed OK (I think), but can't run because it wants visit
(really, it does).
* client installed OK (I think), but can't test without remote

Perhaps unsurprisingly, it's somebody else's code (visit) that causes
the most trouble. Please see my notes and advise on where I went wrong
with that part.

-- 
Michael S. Vogeley
Associate Professor
Director of Graduate Studies
Department of Physics
Drexel University
Philadelphia, PA 19104
(215)895-2710
(215)895-5934 fax
vogeley@drexel.edu
www.physics.drexel.edu

Notes on installing BASIN kernel binaries on mars
-------------------------------------------------
See web page
http://www.physics.drexel.edu/basin_trac/kernel_trac/wiki/BasinInstallationBin

MPICH2
------

Requires mpich2
downloaded
http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/mpich2-1.0.6p1.tar.gz
into ~/Download
become su
cd mpich**
./configure --enable-cxx --enable-mpe --enable-threads=multiple
            by default this should put the executable in /usr/local/bin
make
make install

VISIT
------

Visit is optional, but let's install that, too
Not sure how to parse cryptic comment
"Note: For right now do not attempt to get VisIt working with the binaries!"
I'll ignore this and install Visit 1.6.1

mkdir ~/Download/visit
cd visit

Hmmm. Instructs say to use
wget --no-check-certificate https://wci.llnl.gov/codes/visit/1.7/build_visit
Have to be careful to select 1.6.1 in options during install!
become su
chmod +x build_visit
./build_visit
Follow config instructions on BASIN page...
NOTE - when you select "Version" the second time, it says Version 1.7.
Hit OK, THEN it will let you type in a new version = 1.6.1
After hitting OK on the flags screens, error message
"You must set up the PAR_LIBS to get VisIt to build in parallel.
Stopping build because necessary parallel options are not set."
Looks like it expects the environment variable PAR_LIBS to be set.

My install of mpich2 put the libs in /usr/local/lib,
so how do we reference this?

try
export PAR_LIBS='/usr/local/lib -lmpich'

Now it barfs because it wants PAR_INCLUDE set...
export PAR_INCLUDE='/usr/local/include/mpi.h'

Now the script continues as it should and starts downloading packages...

Dies and complains
"Unable to run. Need to download 1.6.1./visit1.6.1.tar.gz from the
VisIt website, but cannot locate the wget utility
to do so."

Bulll)(*&)^&^. wget lives in /usr/bin/wget. Where exactly is it
expecting it to be? /usr/bin is in the PATH. Let's hack the script by
replacing all instances of wget with /usr/bin/wget and try again...

Nope, that doesn't work, either. Same error message. If I execute the wget command myself from the command line,
wget --no-check-certificate https://wci.llnl/gov/codes/visit/1.6.1/visit1.6.1.tar.gz
I get a 404 Not Found error. Apparently not online anymore.
There IS a file jvisit1.6.1.tar.gz
Is that what they want? Guts of it look right.
Again, wget jvisit1.6.1.tar.gz by hand. THAT exists

OK, now try replacing visit with jvisit in the variable VISIT_VERSION
which labels the tarball in the script so that it will operate on the
file that actually exists.
See my file build_visit3.
Now the script runs past that point and starts to build stuff,
starting with Mesa.
[Note: sure seems like this script was NOT tested by the VisIt folks for versions other than 1.7]
Warning: VisIt takes a long time to build all the components...
Gets to screen saying "Configuring VisIt..."
and then dies with message
"Unable to build or install ViSiT. Bailing out."
See build_visit_log
Problem is with ./build_visit3: line 5986: cd: jvisit1.6.1/src: No such file or directory

Seems that I didn't completely fix the script.
It's looking for a directory that doesn't exist.

Tried gunzipping and untarring jvisit* by hand. It makes src in the
current directory.

Tried changing line 5986 to look in src subdir of working directory.
Now complains that it can't find ./configure.
Grrrr.

***** I quit with visit for now. ************

BASIN KERNEL
-------------

On to the basin kernel itself.

See web page
http://www.physics.drexel.edu/basin_trac/kernel_trac/wiki/Releases

Unsurprisingly,
uname -m
yields
i686

Downloaded basin_kernel_i686_1.2.1-beta_12_13_07.tar.gz to
/home/vogeley/Download
Let's install in /home/vogeley/bin just to be safe for now.
tar -xvf basin_kernel_i686_1.2.1-beta_12_13_07.tar -C /home/vogeley/bin

Next download testcode.C (see bottom of page) into /home/vogeley/bin/basin
Hmm. "save as" or "save link target" on that saves piles of html, not the code.
Perhaps some browser stupidity. Never mind. Just cut and paste into a file.
Edit Makefile to set Source=testcode
Run "make" which executes
mpicxx -I./include -c testcode.C -o testcode.o
mpicxx -L./lib -lbasin -lhdf5 -lcfitsio -ldfftw -ldrfftw -ldfftw_mpi -ldrfftw_mpi -lboost_thread-mt -Wl,--rpath -Wl,./lib testcode.o -o testcode
Seems to have worked OK.
Run ./testcode
which spits out a couple of [0,1] numbers, e.g.,
0.254406
0.474121
Look like the kernel works (at least so far) on mars.
That, unlike the visit nonsense, was easy.

BASIN REMOTE
------------
see http://www.physics.drexel.edu/basin_trac/remote_trac/wiki/RemoteInstallationBin

Download basin_remote_i686_python2.5_1.2.1-beta_12_13_07.tar
from http://www.physics.drexel.edu/basin_trac/remote_trac/wiki/Releases
into my ~/Download directory and unpack
tar -xzf

Which version of setuptools?
I see basin/lib/python2.5/site-packages/setuptools-0.6c7-py2.5.egg

Before I permanently modify my PATH, let's execute
export LD_LIBRARY_PATH=/home/vogeley/bin/basin/lib/qt4:$LD_LIBRARY_PATH
and
export PYTHONPATH=/home/vogeley/bin/basin/lib/python2.5/site-packages/setuptools-0.6c7-py2.5.egg:/home/vogeley/bin/basin/lib/python2.5/site-packages:$PYTHONPATH

Does it work. First run "mpdboot" which yields errors

mpdboot_mars (handle_mpd_output 399): from mpd on mars, invalid port info:
configuration file /home/vogeley/.mpd.conf not found
A file named .mpd.conf file must be present in the user's home
directory (/etc/mpd.conf if root) with read and write access
only for the user, and must contain at least a line with:
MPD_SECRETWORD=<secretword>
One way to safely create this file is to do the following:
  cd $HOME
  touch .mpd.conf
  chmod 600 .mpd.conf
and then use an editor to insert a line like
  MPD_SECRETWORD=mr45-j9z
into the file. (Of course use some other secret word than mr45-j9z.)

Unsurprising, since I've never used mpich on this system.
Follow the instructions in the error message and run "mpdboot" again.
Now start a session with (I'm in the /home/vogeley/bin/basin directory)
scripts/basin_start.sh -n 2 -p 10105
This dies with error messages
== Basin Client Paths ==
Scripts Path: /home/vogeley/bin/basin/scripts
which: no visit in (/usr/local/bin:/usr/bin:/bin:/opt/bin:/usr/i686-pc-linux-gnu/gcc-bin/4.1.1:/usr/i686-pc-linux-gnu/gcc-bin/4.1.1:/opt/blackdown-jdk-1.4.2.03/bin:/opt/blackdown-jdk-1.4.2.03/jre/bin:/home/vogeley/bin:/home/vogeley/xgobi/src)
  VisIt Path:
========================
which: no visit in (/usr/local/bin:/usr/bin:/bin:/opt/bin:/usr/i686-pc-linux-gnu/gcc-bin/4.1.1:/usr/i686-pc-linux-gnu/gcc-bin/4.1.1:/opt/blackdown-jdk-1.4.2.03/bin:/opt/blackdown-jdk-1.4.2.03/jre/bin:/home/vogeley/bin:/home/vogeley/xgobi/src)
scripts/basin_start.sh: line 121: ipcontroller: command not found
Error launching the controller, exiting.

OK, so visit is needed for basin remote? That's not what
http://www.physics.drexel.edu/basin_trac/remote_trac/wiki/RemoteInstallationBin
says.

BASIN CLIENT
------------

Next, the client from binary.
Start at page
http://www.physics.drexel.edu/basin_trac/client_trac/wiki/ClientInstallationBin
From
http://www.physics.drexel.edu/basin_trac/client_trac/wiki/Releases
download
basin_client_i686_python2.5_1.2.1-beta_12_19_07.tar
into my Download directory.
Again, unpack into the /home/vogeley/bin/basin directory
tar -xzf basin_client_i686_python2.5_1.2.1-beta_12_19_07.tar -C /home/vogeley/bin

Will have to wait to get remote installed before testing this.
Received on Mon Jan 7 14:56:42 2008