[ale] Partial Remote Boot

[Chris Fowler]

http://etherboot.sourceforge.net


Check it out.  


-----Original Message-----
From: Jeff Hubbs [mailto:Jhubbs at niit.com]
To: ale at ale.org
Sent: Wednesday, February 21, 2001 4:19 PM
To: ale at ale.org
Subject: [ale] Partial Remote Boot



I'm taking a stab at setting up a MOSIX cluster and I'm going to want the
compute nodes to get some of their OS from a central point (probably via
NFS).  I don't have any boot PROMs, so I expect that the nodes will, at the
very least, boot from floppy or a small HD.  I've read the remote-boot and
diskless HOWTOs but they don't quite tell me what I need.


Specifically, what parts of the filesystem tree should be local and which
could/should be imported, and how can this be set up at install-time given
that I will probably need to use RH 6.2?  I seem to recall that at install
time, I can give an exported file system from another machine a
read/writable mount point (say, /usr) on the target machine and the install
process will fill /usr with files, but what do I do with the SECOND machine?
Do I let it overwrite the exported system, or is there some other trickery?


Let me try to answer my own questions and you can tell me if I'm FOS.


I get the feeling that /lib, /sbin, /bin, and /usr (except for /usr/local)
could be centralized via NFS.  Although I will want every machine to run the
same MOSIX-patched kernel, I know that for some machines the kernel image
should be different (for instance, one of my compute nodes has the buggy
CMD640, so I'll want to have the workaround enabled when I make config for
that particular machine's kernel).  


Going one step further, what if I'm trying to go with a node that's diskless
but boots only from a floppy?  Can I start with tomsrtbt or something even
more cut-down?


- Jeff





--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.