PXEBOOT/TFTPBOOT + big MD_ROOT problem
emmanuel.chriqui at laposte.net
Tue Apr 19 13:12:46 PDT 2005
> Objet : Re: PXEBOOT/TFTPBOOT + big MD_ROOT problem
> Marc Olzheim wrote:
> >On Tue, Apr 19, 2005 at 09:52:07AM +0200, Emmanuel Chriqui wrote:
> >>I'm trying to make very big MD_ROOT (300MB) sent using PXEBOOT+TFTPBOOT.
> >Any reasons for not using NFS ?
> >>I use i386/5.4RC2/TFTPD/PXEBOOT+TFTPBOOT .
> >>(same pb with a 5.3).
> >>Am I missing something obvious?
> >I'm not sure. tftp itself is able to handle 32MB+ files, but maybe the
> >loader isn't.
> >A workaround, no using NFS, could be to tftp a second filesystem image
> >on boot and mount that from the root filesystem.
> I assume that the PXE clients are diskless clients. If so, do they have
> enough memory to handle this extremely large image? As for tfpt, it uses
YES, diskless. And enough memory.
> UDP. UDP is usually used for transfer small datagrams, for instance DNS
> replies. It is also said to be an unreliable protocol. The client should
> repeat the request when no data receives. I doubt this solution is
You're right. However, pros : over 2 years this solution has proven to work
*perfectly* for us, Reboots are rare and done using pools (~10 servers each
time), NFS was horrible to handle and we lost data everytime traffic was
intense. Cons : this was over linux. I believe NFS over FreeBSD works better
(by I still have my linux servers... and they work great so..).
> reliable and flexible enough. My idea for a workaround is creating a
> ramdisk from a small boot image, and transfer the less necessary
> userland binaries from the boot server to the ramdisk using normal ftp
More information about the freebsd-stable