Hello DongInn,

Am Montag, den 14.01.2008, 10:12 -0500 schrieb DongInn Kim:
> Hi Dominik,
> 
> If your openmpi is properly compiled and your modules program works fine but 
> the openmpi test fails because of the switcher issue, then here is my simple 
> therapy.
> 
> 1. First of all, check to see if openmpi was compile with the "tm" support
>    module load openmpi
>    ompi_info | grep tm     # if you see any output about tm here, that is good
> 2. Chroot to image and then setup switcher manually.
>    chroot /var/lib/systemimager/images/oscarimage
>    cd /opt/env-switcher/share/env-switcher/mpi
>    ls  # check to see what mpi switcher is configured.
>        # OSCAR is supposed to configure three mpi names (lam, mpich, and 
> openmpi) by default
>        # If there is anything missing, the corresponding test would fail.
>    cp /opt/openmpi-switcher-modulefile-1.2.4/share/openmpi/openmpi-1.2.4 . # 
> Assuming openmpi name is missing.
> 3. Once the image is updated, reimage your client nodes with the new image
> 4. Or not recommend but cpush the missing mpi names to the client nodes.
>    cd /opt/env-switcher/share/env-switcher/mpi
>    cpush openmpi-1.2.4

Thank you. I try this today if it works with your step by step manual.

> BTW, what kind of error did you get when you compiled modules-oscar? What did 
> you change on the modules-oscar.spec file?

I got this output at step 8:

Performing root tests...
Maui service check:maui                                        [PASSED]
TORQUE node check                                              [PASSED]
TORQUE service check:pbs_server                                [PASSED]
/home mounts                                                   [PASSED]

Preparing user tests...
Performing user tests...
mpi/openmpi-1.1.1(64):ERROR:102: Tcl command execution failed: if
{ $have_blcr =
= 1 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-blcr-and-gm-$version] } {
  module load $dir/openmpi-with-blcr-and-gm-$version
} elseif { $have_blcr == 0 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-gm-$version] } {
  module load $dir/openmpi-with-gm-$version
} elseif { $have_blcr == 1 && $have_gm == 0 &&
     [file exists $dir/openmpi-with-blcr-$version] } {
  module load $dir/openmpi-with-blcr-$version
} elseif { [file exists $dir/openmpi-$version] } {
  module load $dir/openmpi-$version
} else {
  break
}

SSH ping test                                                  [PASSED]
SSH server->node                                               [PASSED]
SSH node->server                                               [PASSED]
mpi/openmpi-1.1.1(64):ERROR:102: Tcl command execution failed: if
{ $have_blcr =
= 1 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-blcr-and-gm-$version] } {
  module load $dir/openmpi-with-blcr-and-gm-$version
} elseif { $have_blcr == 0 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-gm-$version] } {
  module load $dir/openmpi-with-gm-$version
} elseif { $have_blcr == 1 && $have_gm == 0 &&
     [file exists $dir/openmpi-with-blcr-$version] } {
  module load $dir/openmpi-with-blcr-$version
} elseif { [file exists $dir/openmpi-$version] } {
  module load $dir/openmpi-$version
} else {
  break
}

ModuleCmd_Load.c(199):ERROR:105: Unable to locate a modulefile for
'switcher'
Checking for 2 free nodes:                                     [FAILED]
Not enough free nodes. Tests incomplete.
mpi/openmpi-1.1.1(64):ERROR:102: Tcl command execution failed: if
{ $have_blcr =
= 1 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-blcr-and-gm-$version] } {
  module load $dir/openmpi-with-blcr-and-gm-$version
} elseif { $have_blcr == 0 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-gm-$version] } {
  module load $dir/openmpi-with-gm-$version
} elseif { $have_blcr == 1 && $have_gm == 0 &&
     [file exists $dir/openmpi-with-blcr-$version] } {
  module load $dir/openmpi-with-blcr-$version
} elseif { [file exists $dir/openmpi-$version] } {
  module load $dir/openmpi-$version
} else {
  break
}

TORQUE default queue definition                                [PASSED]
Checking for 2 free nodes:                                     [FAILED]
Not enough free nodes. Tests incomplete.
mpi/openmpi-1.1.1(64):ERROR:102: Tcl command execution failed: if
{ $have_blcr =
= 1 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-blcr-and-gm-$version] } {
  module load $dir/openmpi-with-blcr-and-gm-$version
} elseif { $have_blcr == 0 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-gm-$version] } {
  module load $dir/openmpi-with-gm-$version
} elseif { $have_blcr == 1 && $have_gm == 0 &&
     [file exists $dir/openmpi-with-blcr-$version] } {
  module load $dir/openmpi-with-blcr-$version
} elseif { [file exists $dir/openmpi-$version] } {
  module load $dir/openmpi-$version
} else {
  break
}

ModuleCmd_Load.c(199):ERROR:105: Unable to locate a modulefile for
'switcher'
Checking for 2 free nodes:                                     [FAILED]
Not enough free nodes. Tests incomplete.

Run APItests...

Running Installation tests for pvm
[PASS]       2008-01-11 15:53:29   pvmd-path-ls.apt
[PASS]       2008-01-11 15:53:29   envvar-pvm_arch.apt
[PASS]       2008-01-11 15:53:29   envvar-pvm_root.apt
[PASS]       2008-01-11 15:53:29   envvar.apb
[PASS]       2008-01-11 15:53:29   pvmd-path-which.apt
[PASS]       2008-01-11 15:53:30   modulecmd-path-ls.apt
[PASS]       2008-01-11 15:53:30   pvm-module-list.apt
[PASS]       2008-01-11 15:53:30   pvm-module-show-pvm_rsh.apt
[PASS]       2008-01-11 15:53:30   pvm-module-show-pvm_arch.apt
[PASS]       2008-01-11 15:53:30   pvm-module-show-pvm_root.apt
[PASS]       2008-01-11 15:53:30   pvm-module-show.apb
[PASS]       2008-01-11 15:53:30   pvm-module.apb
[PASS]       2008-01-11 15:53:30   install_tests.apb

There are 5 failed/skipped tests (see above).
Please check for .err and .out files in /home/oscartst/<package>.

...Hit <ENTER> to close this window...


I get rid of the

mpi/openmpi-1.1.1(64):ERROR:102: Tcl command execution failed: if
{ $have_blcr =
= 1 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-blcr-and-gm-$version] } {
  module load $dir/openmpi-with-blcr-and-gm-$version
} elseif { $have_blcr == 0 && $have_gm == 1 &&
     [file exists $dir/openmpi-with-gm-$version] } {
  module load $dir/openmpi-with-gm-$version
} elseif { $have_blcr == 1 && $have_gm == 0 &&
     [file exists $dir/openmpi-with-blcr-$version] } {
  module load $dir/openmpi-with-blcr-$version
} elseif { [file exists $dir/openmpi-$version] } {
  module load $dir/openmpi-$version
} else {
  break
}

stuff by adding just the aaa_base requirement to the modules-oscar.spec
like I did at the openSUSE 10.2 (OSCAR 5.0; x86_64) work I did.
But without testing openmpi I can not say if the problems are definitiv
fixed. I'll do some more tests today.

-- 
Mit freundlichen Grüßen / Best regards

Dominik Schips

Tel.: +49 (0)21 61 - 46 43-112
Fax:  +49 (0)21 61 - 46 43-100

credativ GmbH, HRB Mönchengladbach 12080
Hohenzollernstr. 133, 41061 Mönchengladbach
Geschäftsführung: Dr. Michael Meskes, Jörg Folz


-------------------------------------------------------------------------
Check out the new SourceForge.net Marketplace.
It's the best place to buy or sell services for
just about anything Open Source.
http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace
_______________________________________________
Oscar-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/oscar-devel

Reply via email to