Hermes Software Configuration Changes in 2018, Part 2

Links below were the guides to what was done. It should be assumed that the indicated steps were executed by root.

More files from Perseus were recovered... And additional steps were performed in September, starting on 9/5/18.

But prior to recovering more files, on 8/7/18, Mathematica 11.3 was installed by running the Wolfram script and saying "yes" a lot.

Some of the old Perseus HDDs had not yet been copied over to /data2; the remaining files were taken care of during September.
One of the old Perseus drives was left in the available drive bay in Hermes. It was given the mount point /data3 in /etc/fstab.

Before doing another "yum update", it seemed like it would be good to restart the machine and bring it back to a clean state.
However, the restart stalled at POST and would go no further. This was because the old /home drive from Perseus was still present.
Previously this drive had been hot-swapped in and live-mounted; it was the old main boot disk, so it would cause problems at boot time.
Solution: shut down, remove the disk, reboot into emergency mode, and edit /etc/fstab so the disk wouldn't be missed at next boot.
After that, "yum install" proceeded smoothly with no issues.

To see if this was enough to clear up the ParaView issues, ParaView 5.5.2 was downloaded, but neither it nor 5.4.0 would run properly.
In the console, both complained about OpenGL not being recent enough. In X2Go, ParaView simply dumped core.

UPDATE: This issue was resolved when ParaView 5.8.1 was installed during the 2020 system refresh.

One possible hint might appear in the output of the following command:

grep NVIDIA /var/log/Xorg.0.log
[ 16171.935] (II) Module nvidia: vendor="NVIDIA Corporation"
[ 16171.935] (II) NVIDIA dlloader X Driver  390.25  Wed Jan 24 18:57:05 PST 2018
...
[ 16171.940] (EE) NVIDIA(0): Failed to initialize the GLX module; please check in your X
[ 16171.940] (EE) NVIDIA(0):     log file that the GLX module has been loaded in your X
[ 16171.940] (EE) NVIDIA(0):     server, and that the module is the NVIDIA GLX module.  If
[ 16171.940] (EE) NVIDIA(0):     you continue to encounter problems, Please try
[ 16171.940] (EE) NVIDIA(0):     reinstalling the NVIDIA driver.
...

In the console, ParaView 5.4.0 and 5.5.2 stated that the OpenGL driver was Mesa; either Mesa or the NVIDIA drivers could be to blame.
Updating the NVIDIA drivers seemed like the logical next step. It is generally a good idea to update the CUDA driver at the same time.
Usually the CUDA driver and toolkit should be updated first, as doing this typically installs an older version of the NVIDIA driver.
However, in this case the CUDA driver installation included a newer NVIDIA driver, so the separate NVIDIA driver installation was skipped.
Following a reboot, all ParaView versions started to work correctly, but only on the console.

A further OS update was run around this time, to ensure that all graphics-related software was in the best possible state.
It didn't help ParaView, and the update to microcode_ctl 2.2.1 (perhaps related to Spectre/Meltdown?) may even have been a step backward.

The Intel Fortran compiler was updated... The 2018 and 2019 releases were both added in October, on 10/2/18.

The license for the Intel Fortran compiler on Hermes included one year of support, which was set to expire on 10/7/18.
Luckily, version 19.0.0 was released just days ahead of the deadline, so it made sense to install this version before support expired.
Since 19.0.0 was so new, both it and 18.0.3 were installed in standard locations in /opt/intel, alongside the previously installed 17.0.4.
Version 19.0.0 became the new default because it was the last one installed. This was verified by typing "mpif90 --version".



Last updated on 7/30/19 by Steve Lantz (steve.lantz ~at~ cornell.edu)