Issues installing Libmesh on cluster

[TheGreatCid]

Hi All,

I am installing Libmesh on a cluster with the following modules activated.

cmake/3.19.3-rhel8
Python/3.8.1
GCC/9.3.0
MPICH/3.2.1
OpenMPI/4.0.5-rhel8

I am able to install Petsc without any issues.
To install libmesh I am using the update_and_rebuild_libmesh.sh script that comes with MOOSE.
I am able to configure libmesh, but the script fails when it tries to run make install
I receive the following error:

Making install in include
make[1]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
Making install in libmesh
make[2]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make  install-am
make[3]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make[4]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make[4]: Nothing to be done for 'install-exec-am'.
make[4]: Nothing to be done for 'install-data-am'.
make[4]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make[3]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make[2]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include/libmesh'
make[2]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
make[3]: Entering directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
make[3]: Nothing to be done for 'install-exec-am'.
 /usr/bin/mkdir -p '/usr/local/include/libmesh'
/usr/bin/mkdir: cannot create directory ‘/usr/local/include/libmesh’: Permission denied
make[3]: *** [Makefile:1150: install-includeHEADERS] Error 1
make[3]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
make[2]: *** [Makefile:1344: install-am] Error 2
make[2]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
make[1]: *** [Makefile:1177: install-recursive] Error 1
make[1]: Leaving directory '/hpc/group/dolbowlab/det12/raccoon/moose/libmesh/build/include'
make: *** [Makefile:32937: install-recursive] Error 1

Libmesh is trying to write in a directory that I do not have access to.
Is there any way to avoid this?
I have run this script successfully on this cluster before, so I am unsure of what has changed in how the script executes.

Thank you in advance!

[GiudGiud]

Hello

So first you should only have one MPI library loaded

MPICH/3.2.1
OpenMPI/4.0.5-rhel8

are both MPIs

This is odd. What does env return? We're looking for an environment variable that is changing the install path

Guillaume

[roystgnr]

Wait - did you say the whole config.log file appears cut off at that point, not just the configure line?

If something (full filesystem, process killed, I dunno) somehow managed to kill a configure run right in the middle of writing out that line, that would be a one-in-a-billion "inexplicable went awry" event, but it might explain what's going on since.

[TheGreatCid]

No, I misspoke, apologies. I just meant the configure line.

I reset my branch to head, but now I do remember that I did modify the config file
If I recall correctly, I removed the line that requires xdr.
I must have done it incorrectly, seeing that is where the options stop being read...
I believe that must be the issue.
My apologies for all of this

[roystgnr]

For future reference, if you need to override a single configure argument, our scripts take any arguments they don't recognize and append them to the configure line, and when autotools sees conflicting arguments it follows the rule "later argument wins", so I think you can just update_and_rebuild_libmesh.sh --disable-xdr-required to turn that off.

(though I'd recommend not doing so with that particular feature, if there's any way you can avoid it - although IMHO .xda.gz and .cpa.gz are just as good as .xdr and .cpr, a lot of code has been written which disagrees with me, and the number of MOOSE tests that fail when they can't even read/write a binary restart file isn't pretty)

[milljm]

I was about to ask for a git diff in your moose directory. That would have illuminated the issue. But then I thought by asking for a git diff, that would have been saying something rude "I don't believe you". Lesson learned!

[TheGreatCid]

though I'd recommend not doing so with that particular feature, if there's any way you can avoid it

I'll see if I can leave it enabled. It has given me issues in the past, which is why I was disabling it.

But then I thought by asking for a git diff, that would have been saying something rude

I would not have seen that as rude, but I suppose I can't speak for everyone
I should have done that to begin with instead of trusting my memory...