[sf-lug] Server Installs in large environments

[Mike]

Hey folks,

At last Sunday's meeting I was told this would be a good place to post
this problem, so here it goes:

I work in shop that has 120+ production servers, on OpenSuSE, and 
increasing in number. 

Currently we build our servers from either a usb disk or sata drive,
installed directly on/in the box, and use a python script to do a bare
metal recovery based around dump, parted, fsck, etc.  It is labor
intensive, and with the expectation to double in size next year, is
really not a scalable solution.

In our environment, we have 32 bit 1U application servers, 64 bit 1U
application servers, and 16 drive 64 bit 3U database servers. about
5-10% of our servers require a custom setup; we use most of our 1U's
interchangeably, and all the 64 bit 1U systems have the same drives. 
Not sure about how varied the 32 bit 1U drives are, but as they go out
of warranty, they are pulled from production and replaced with 64 bit
1U's.  Our 64 bit 3U's have two distinct drive configurations, and as
they go out of warranty, they too, are pulled from production.

I am aware of network-based solutions such as autoYaST/Kickstart and I
have read about System Imager, but I would like to hear about what has
been tried successfully and unsuccessfully in large environments as I
have no experience in this area, never managing more that 6-7 servers at
a time before starting this job.  The idea of managing one image, that
can either be installed over a network, and/or from a DVD, followed by a
post install configuration script depending on server type, is highly
appealing, but I also know that may be unrealistic.

Thanks in advance, and let me know if you require additional info.

Mike