Filesystem Hierarchy Standard Respecting CentOS

Home » CentOS » Filesystem Hierarchy Standard Respecting CentOS
CentOS 1 Comment

I am investigating using rpm to package one of our in house applications. This software started life as a sub-system within a Ruby-on-Rails application but has now been extracted into its own standalone package, none of which has any httpd access.

The local package under consideration will run as a set of cron jobs under a designated userid. There will be no user interaction. A
remote database server will be updated but no data files will remain on the installation host once they are processed.

I have read the subject document a couple of times and I am wondering what people here with experience in these matters actually do. The FHS suggests to me that projects of the sort I am dealing with belong entirely within the /opt/package_name/ tree with variable data stowed in /var/opt/package_name or possibly /var/spool/package_name.

However, I have not yet found any application packages for CentOS-6
that actually do this. I find some that go into /usr/package_name, some into /usr/lib/package_name, many that install into /usr/libexec and none that install into /usr/local, which I gather is reserved for packaged built on the system rather than installed via rpm.

So, what is the actual practice with respect to packaging via rpm?
Where do things go?

One thought on - Filesystem Hierarchy Standard Respecting CentOS

  • I’ve seen a few. Not many, but a few. (some IBM products, a couple of other commercial products).

    It’s reserved for the local admins to use; they _could_ do their own rpms that install into /usr/local, but no third party rpm should touch it.

    I’d recommend following the FHS standard locations. If you ever decide to support more than just CentOS then all Linux distro’s should “support”
    that structure, as do other non-Linux based Unix systems.