What you call different Applications for staff and OPAC would be implemented as two Apache VirtualHosts. They share a lot of the underlying perl code in common, so it it usually more efficient to run them on the same server. On Mon, Apr 6, 2009 at 3:48 PM, Erik Lewis <elewis@ngrl.org> wrote:

I understand that Koha is essentially 2 parts an Apache application server and a MySQL database server. I also understand that some libraries run a seperate application server for staff access and one for patron access. If I'm not mistaken you can run these on different machines as well.

1) Is there any problem mixing Application servers with the Database server. Example a Mac serving MySQL and either Windows, Mac, or Linux App servers?

The mysql server can be whatever, but I would stick to one of the major Linux distros for the Apache/Koha server(s) unless you have a very strong reason not to, and are happy about debugging as you go. The system-dependent chunks are in Koha's compiled perl module dependencies, so mysqld is unaffected. Some folks do use MacOS for serving Koha though.

2) Any upper limit on the number of App servers?

No, but it isn't really necessary to build a huge "farm" either.

3) Where does the Zebra indexing live?

Wherever. It can even be a 3rd separate server, talking to the same DB server, with it's own copy of the Koha code, or a mounted drive of the same dirs. Keeping code in sync during updates can be a problem with this approach, however. I'd say keep Zebra on your Apache server unless you find performance issues.

4) What are the pitfalls of having the Database server on a separate server?

Firewalls, and more points of hardware failure. Introduces network performance/reliability bottleneck for every operation. Complexity of administration (OS updated, security, etc.). --Joe