It should work, but you will need to do some planning around doing updates on the master copy (Bill Scully's paper on how CA does this is instructive as to the various issues).
One thing that I've been tinkering with is whether this sort of configuration is really more like setting up diskless clients in the discrete world. If one were to set up a pair of guest LANs (one as "front" side for public traffic, one for back-end between-server-only communications), attach a "boot" host to the backside gLAN providing TFTP, DHCP and NFS service, and several "disk" server guests to export chunks of writable space via NFS to specific clients, and then define each guest with two interfaces with specific fixed MAC addresses, I think it would be possible to do all the configuration and shared disk management via the same methods as are used for diskless clients in the discrete world. The boot server would handle IP address assignment, provision of root file systems for each client, and the "disk" servers could handle distributing private space to individual servers via whatever technology needed. Updates could be done trivially by chroot'ing to a particular server's subset of the disk space and then running the install tools as if on the individual server, and the use of NFS to coordinate I/O would avoid the caching of blocks in multiple virtual machines problem that physically shared disk exhibits. Creating a client would then be running a script on the boot server to associate a MAC address with a particular host configuration, allocating and creating a subtree in /export/root/<client> and populating it, and allocating a single shared disk with just enough code on it to boot and initialize the network interfaces and get boot info via DHCP/bootp. Everything else would come over the network (which would be OK, as most of the issues with network clients are caused by physical network failures, which a GLAN isn't subject to -- or if it does fail under z/VM, you're already horked beyond repair). Solaris has been doing this kind of thing for several decades now, and it's proved to be a useful thing in a lot of different configurations. I'll be publishing some results soonish, and the scripts will be made available if this turns out to amount to something useful. Definitely a WAVV presentation for next year, though -- it's proving to be an interesting problem. -- db ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
