[PL #7599] Re: PlanetLab diskless node project

[Marc E. Fiuczynski via RT]

Email Recipients (see http://www.planet-lab.org/Support)
       Owner: Nobody
       Requestor: alsbergt at cs.huji.ac.il
       Ticket Ccs: alk at absarokasoft.com, daniel51 at cs.huji.ac.il, mef at cs.princeton.edu, smuir at cs.princeton.edu

==================================================

> From this experience I understood that the BootManager is not a small
> boot script but a modular program handling many important tasks of
> setting up, installing the node, and communicating with the PLC.  I
> now see, as you immediately said, that trying to bypass the
> BootManager is not a good idea, and will try to make all changes for
> the diskless project to fit well into the BootManager design notes and
> for being well integrated with other non-diskless nodes and the PLC.  

Good.

> P.S. As for your mail about the NFS vservers patch, I will have to
> still look at what this patch is about exactly.  I suppose there will
> be no problem in making the NFS server be a Linux host with that patch
> as well (although from sysadmin experience Linux's NFS implementation
> is rather buggy and I usually prefer fileservers to run FreeBSD -- or
> some other system better suited for that, like Solaris or some NetApp
> -- those are problems that can be overcome later if they appear and
> need not be worried about now).

vservers uses some unused fields within the NFS protocol to differentiate between different vserver context.  This wont work with a FreeBSD, Solaris, or NetApp NFS server.  You'll have to stick with one based on Linux + vserver patches.

Marc