[comp.sys.sgi] Sun/SGI boot interference

cunning@arthur.msd.lmsc.lockheed.com (Dave Cunningham) (09/12/90)

I'm being told by another site at my company that my Irises are
causing problems that prevent thier diskless suns from booting.  The
suns apparently loose track of what YP domain they are part of
somewhere during the boot process and hang.  Folks at the other site
believe, based on some past experience that I haven't been able to get
a detailed report on, that the problem is a bug in the Iris's
bootparamd.  Has anyone else out there experienced this ?

Configuration details are:

  Suns running SunOs 4.0.3, Irises 3.2.1
  Both sites on a single (bridged) ethernet
  Irises are diskfull, suns are diskless clients and a sun server
  Both sites run nfs and yp;  different yp (ooops, nis) domains

The fix I'm being asked to apply is to comment out the bootparamd and
bootp lines in my inetd.conf files on the Irises.  I don't boot the
irises diskless, but I do install software across the net; I believe
that bootp, at least, is necessary to get sash running.

If anyone has any pertinent info or experience I'd appreciate a note.
Thanks in advance,


Dave Cunningham				cunning@arthur.msd.lmsc.lockheed.com
Lockheed Missiles and Space Co.
O/81-10, B/157-5E, (408)-756-1382

#include  <standard_disclaimer.h>

phil@BRL.MIL (Phil Dykstra) (09/13/90)

We have experienced the "Sun booting problem" as well.  Many versions
of bootparamd answer requests even if they have no file systems for
the requesting machine.  This causes a race condition where a SunOS 4.x
machine will try to mount file systems from the first machine to answer
its broadcast bootparam request.  We commented out bootparm in
/usr/etc/inetd.conf on all of our SGI's.  All versions of Irix 3.2
seem to have this problem [as well as some other computers like the
TI Lisp Machines].  Irix 3.3 may have fixed it - I don't know.

We did not have to do anything involving bootp.

- Phil

blbates@AERO4.LARC.NASA.GOV ("Brent L. Bates AAD/TAB MS361 x42854") (09/13/90)

    I think I have heard of that bug before, and it is fixed in IRIX 3.3.
--

	Brent L. Bates
	NASA-Langley Research Center
	M.S. 361
	Hampton, Virginia  23665-5225
	(804) 864-2854
	E-mail: blbates@aero4.larc.nasa.gov or blbates@aero2.larc.nasa.gov

cunning@arthur.msd.lmsc.lockheed.com (Dave Cunningham) (09/13/90)

Thank you to everyone who replied to my question about Irises interfering with
diskless Sun's boot procedures.  The concensus answer was

	Yes, 3.2 (at least) Irises have a misbehaving bootparamd which
	interferes with 4.0.3 (at least) Suns trying to boot from a server.
	Lots of sites have run into the problem.

	Commenting out the bootparam line in /usr/etc/inetd.conf on the
	Irises fixes the problem.  (Unless, of course you need to boot
	Irises diskless...).  It is NOT necessary to disable the bootp
	server.  You also need to reboot or send a HUP to inetd, and
	possibly kill a lingering process running rpc.boot.

	The problem is supposed to be fixed in Irix 3.2.x, x>1,
	and in 3.3.  There is also a SunOs patch for 4.0.3 which may
	avoid the problem; it's call the "jumbo nfs" patch.  Don't know
	about SunOs 4.1.

Thanks again for the helpful responses.

#include  <standard_disclaimer.h>

Dave Cunningham				cunning@arthur.msd.lmsc.lockheed.com
Lockheed Missiles and Space Co.
O/81-10, B/157-5E, (408)-756-1382