[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Python and Condor

My two cents.

In fact I got python 2.3.6 compiled with condor libs and it seems to
work. However, I need Numeric module and here I failed to get it

For python, I am doing this:

## For Python 2.3.6
cd; tar xvfj Python-2.3.6.tar.bz2
cd Python-2.3.6
make clean
./configure --prefix=$HOME/PythonCondor --without-thread
condor_compile gcc -Xlinker -export-dynamic -o python Modules/python.o
libpython2.3.a -ldl -lutil -lm
make install

I didn't test with Python 2.4, and with 2.5 it failed when I do 'make install'

When trying to condor_compile Numeric, I got the famous "Internal ld
was not invoked!". BTW, 'condor_compile make' for Python above will
probably give the same error. Frustrating sometimes why condor_compile
works and (most of time) not.

Good luck.
PS:  If someone has any other idea what I can do to go further I am
prone to test.

On 29/03/07, Nick LeRoy <nleroy@xxxxxxxxxxx> wrote:
On Thu March 29 2007, simon kagwe wrote:
> Hi,
>   Thank you Nick for your answers.
>   So does it mean the machines will have to share a filesystem.? I was
> trying to avoid using NFS or AFS. The whole reason I was going to implement
> it on Linux was to take advantage of checkpointing and remote system calls.

No, not at all.  It just means that you need to build your own Python
interpreter on your submit machine (let's say you've installed it
as /usr/local/bin/cpython).  Then, when you submit your job, your submit file
would look something like this

universe = standard
executable = /usr/local/bin/cpython
arguments = myscript.py
should_transfer_files = yes
when_to_transfer_output = ON_EXIT
transfer_executable = true
transfer_input_files = myscript.py

Condor will transfer the executable for you (because transfer_executable is
true - you shouldn't need to specify this, but it doesn't hurt), and your
script (because it's in the input file list).

Also, if my memory serves me correctly, people have built standard universe
pythons before.

All warranties, express and implied, are void.  ;)

Hope this helps