On Mon, 30 Aug 2021 11:41:31 +0000, g4sra wrote in message
<mdveoGC21RVu_O2Qo5iINsVtEiDRenBEyPtteCKBHsN2KQhj7eTwsWw_BmpyubRbH1BPxfU06mPG16DAlhYwBUfVaolss8kZMxmjYqpdMTY=@protonmail.com>:

> On Monday, August 30th, 2021 at 12:10 PM, Luciano Mannucci
> <luciano@???> wrote:
>
> > Hello,
> > I have two Beowulf systems that should be identical. If I issue "df"
> > one does report among mounted filesystems a /run/user/xxx, where xxx
> > is the id number of the user I am connected with (via ssh) and the
> > other doesn't. As uname -a they both report:
> > Linux mil-dbs66 4.19.0-14-amd64 #1 SMP Debian 4.19.171-2
> > (2021-01-30) x86_64 GNU/Linux What have I done wrong, and on which
> > one of the two machines?
> Not necessarily anything wrong, other than the two machines are not
> identical.
>
> /run/user/xxx is used by session management (and in some corner cases
> its absence can break dpkg scripts).
>
> You could list all the processes on both machines and then diff them.
> Quick and dirty (and untested :)
>
> local_host# ps -e | awk '{print $4}' | sort | uniq >/tmp/catch0
> local_host# ssh remote_machine 'ps -e' | awk '{print $4}' | sort |
> uniq >/tmp/catch1 local_host# diff /tmp/catch0 /tmp/catch1

..a "dpkg --get-selections >>/tmp/catch$B" on each Beowulf might tell
you why your Beowulf systems act differently. If not, you'll have to
e.g. diff both /etc etc trees, but you shouldn't have to take it as
far as diff'ing both / trees.

--
..med vennlig hilsen = with Kind Regards from Arnt Karlsen
...with a number of polar bear hunters in his ancestry...
Scenarios always come in sets of three:
best case, worst case, and just in case.