Re: How to install a bunch of RPMS if some are already installed?

[Alan Hagge <ahagge wbfa com>]

Thanks for the quick responses.  Actually, I'd already figured out that 
running through a loop on every package with a "rpm -U"  WOULD obtain 
the desired results (with messages printed for all of the stuff already 
installed, of course, and assuming there are no interdependencies 
between the packages), but what I was HOPING for was a single command 
which would perform the intended operation, as doing a lot of separate 
calls involves a LOT of duplicated overhead (executing rpm, opening the 
database, comparisons, etc.).  It takes about 10 times as long, even 
with < 100 files.

I just found it hard to believe that no one else is doing upgrades this way:

   * Create a directory containing all of the "add-on" packages that
     you want every machine to have (in addition to their "standard"
     install).  This would contain 3rd-party apps not incorporated into
     the RedHat distros, etc.
   * Once a (day|week|etc.) run an rpm command on each client to ensure
     that it was up-to-date with any changes to this "add-on" list.
      Then pushing new software out to all of your systems becomes
     simply a matter of dropping the rpm into this directory.  VERY
     handy for administering a large group of machines.

Any chance of making this a "feature request" for the next version of 
RPM, possibly via an option to "rpm -U", say "-k, --keep" (a la sgi's 
"inst" command's "keep" option)?

Thanks again,

Alan

> Jag <agrajag@dragaera.net> wrote:
>
>  
>
> Probably a simple for loop would sort them out for you.  Then you
> could remove those that are dubs (or find other problems, but still
> get installable ones installed)
>
> I haven't done this but I think the rpm --test command may be your
> friend here.  I only tried agains a very small sample.
>
> Probably work without the --test too, but testing  would be a little safer.
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>  #!/bin/bash
>  RPM_DIR=/some/rpm_dir
>  ## cd to the scene of the crime
>  cd $RPM_DIR
>
>  for ii in `ls *rpm`  ## get a working list of all the rpms in there
>  do
>    ## test and record (including stderr)  each file
>    rpm -Uvh --test $ii > test_results 2>&1
>    ## if test shows exit status of 0 then run the install command.
>    if [ "$? = 0 ]; then
>       ## record installationn 
>       rpm -Uvh $ii 2>&1|tee -a installed_results 
>    fi
>  done
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>
> That should leave you with the installable ones installed,
> uninstallable ones will be recorded in test_results
> In case there were other reasons besides being installed already.
>
> Install output will be recorded in installed_results but also piped
> to screen so you see what is happening.
>