You are being much too vague for useful response to be generated.
for example, give an example of an rpm that has the problems that you
are talking about. I suspect that it will be an rpm that would be better
installed to each system anyways.
typically a distribution is installed to each client. all 3rd parrt s/w is
installed to either /usr/local or /opt, and the packager generally realizes
that they are 3rd party software, and will usually configure themselves
to /opt
if the above is properly done, then the server exports /opt and/or /usr/local
and the cliens all nfs mount /opt and/or /usr/local
stuff in /usr, /lib, /etc, etc, are not meant to be exported.
rgds,
-Greg
On 23-Oct-00 Michael Jinks wrote:
> Greetings.
>
> RPM and NFS are both great tools for simplifying network administration, but
> I see an apparent conflict and I'm not sure what the "standard" approach is.
>
> Suppose I want to install an rpm package for use on the network. A simple
> "rpm -ivh 'packagename'" on the NFS server isn't such a good idea, as the
> package will be stored within the server's main file system, which might not
> be (isn't) the same filesystem which is exported for use by the clients.
>
> Okay, so I create (say) /export/usr, and store my binaries there. Trouble
> with that is that rpm doesn't just write binaries, it also writes config
> files, library files, and initial versions of variable data.
>
> Okay, so I create (say) /export, and install an entire Linux system there,
> including /export/var, /export/lib, /export/usr/lib, /export/etc, and all the
> rest of the files that would be necessary to run a full-blast Linux system.
> Trouble now is that some files do actually need to be local to the machine
> where the software will run, for example state information that might be
> written to /var.
>
> I'm also a bit confused about how to get a given package to look in the right
> place for its config files. If the package was configured to get config info
> from /etc/foo.conf, there's no systematic, generalized way to tell a class of
> packages "prepend $PATH to all config files, thus /etc/foo.conf becomes
> $PATH/etc/foo.conf". Even the --root directive to rpm -i doesn't take care
> of this if the path as written on the server differs from the path as read by
> the client -- which it almost always does in my experience.
>
> Was RPM just not designed with NFS-heavy networks in mind? Or am I missing
> something basic? At this point it seems as though I'm going to be stuck
> visiting (or automating the visitation of) every machine on my network, and
> running an rpm command on each one every time we want to change our software
> package loadout.
>
> I'm hoping that somebody who does this can suggest some tricks.
>
> Thanks,
> -m
>
> --
> Michael Jinks, IB
> Systems Administrator, Saecos Corporation
> HOORAY FOR POKEY!!! http://www.yellow5.com/pokey/
>
>
>
> _______________________________________________
> Redhat-list mailing list
> [EMAIL PROTECTED]
> https://listman.redhat.com/mailman/listinfo/redhat-list
----------------------------------
E-Mail: Gregory Hosler <[EMAIL PROTECTED]>
Date: 24-Oct-00
Time: 10:21:17
If Bill Gates had a dime for every time a Windows box crashed...
...Oh, wait a minute, he already does.
----------------------------------
_______________________________________________
Redhat-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list
