Hi Michael, Thanks for your detailed reply. Very helpful.
What if I build openpkg with : $ sh openpkg-***-***.src.sh --prefix=/ --user=here --group=here Will that break anything ? Though I appreciate the ability to run multiple instances of an app on one box, (This is one thing sadly missing in any other packaging system ... and I've researched quite a few) my problem is that we run an ASP environment and changing application paths at this point is just not an option. As far as deployment is concerned, I currently use an overly infrastructure over ssh/scp to manage applications using push. I have a good mind to move to using cfengine, but there are some timing issues we faced while testing that. Thanks again, -ansh -----Original Message----- From: Michael Schloh von Bennewitz [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 05, 2003 9:05 AM To: [EMAIL PROTECTED] Subject: Re: Trying to understand openpkg installroot Hi Anshuman, On Tue, Nov 04, 2003, Anshuman Kanwar wrote: > I've read through the handbook and other docs on the openpkg site. From what > I understand, openpkg will install all packages under its own installroot ? > Is this true ? Is there a way to install these packages in the system path ? > After you bootstrap an OpenPKG instance (running sh openpkg-***-***.src.sh), all packages built from the 'rpm' in that OpenPKG instance will belong to that instance. OpenPKG does not offer a way to build a 'relocatable' binary package. However, it is a packaging system so self contained that you can bootstrap multiple instances on a single machine. To easily use the tools you install in each instance, you can add the respective '/opkg_pfx/bin' to your path. > I want to build my own packages on one machine and deliver on a fleet of 400 > identical machines. In essence I just want a package delivery system ... the > default openpkg way seems too intrusive. Could someone explain why openpkg > needs to maintain its own self contained subtree ? > OpenPKG does not have any built in features to deploy onto a network or check the success of its operations, but in some cases one can write a script to handle deployment. I'm wondering how you presently deploy and manage software on 400 machines. OpenPKG enforces its containment so that files don't conflict or write over each other without the admin's knowlege. If the config files of a ftp daemon package were allowed to install in /etc, then it would mean that only a single such package could be installed on the machine. We have machines with up to 15 instances on them (though most have just a few). It sounds like you want just one identical OpenPKG instance on each of your 400 machines. What about a script on each machine like: #! /bin/sh $ sh openpkg-***-***.src.sh --prefix=/here --user=here --group=here $ sudo sh openpkg-***-***.arch.sh $ /here/bin/rpm --rebuild ftp://ftp.openpkg.org/release/1.3/SRC/openpkg-tool* $ sudo /here/bin/rpm -Uvh /path/to/openpkg-tool.arch.rpm $ /here/bin/openpkg build -p sudo gcc flex bison sendmail | sh Would you like a special arrangement of where the installed files go? Can you explain your wishes a little more? Regards, Michael -- [EMAIL PROTECTED] Development Team, Operations Northern Europe Cable & Wireless Telecommunications Services GmbH ______________________________________________________________________ The OpenPKG Project www.openpkg.org User Communication List [EMAIL PROTECTED]
