Re: (fwd) 2 Pb with Etherboot

[markus@infoscape.com]

[ I forwarded this message to the mailing list, hoping that someone
  else might have experienced similar problems.
  When responding, please cc Denis, because he is currently not (yet)
  subscribe to the list. ]

denis writes:
 > 1- The "disk boot prom all.com" is not able to find 
 > the server if the server never did initiate any request 
 > on the net from boot. 
 > This is not a hardwar problem (see Rq1 and Rq2).
 > 
 > Rq1 : This is not true (the server is responding
 > correctly), if i boot a linux client with 
 > a local disk and try a bootptest and a tftp transfer. 
 > 
 > Rq2 : If i initiate a frist acces by any method
 > and do not reset the server, the client will find the server.
 > (Boot a linux client and tftp or bootptest, ping on the serveur a
 > non present address, ...)

At first sight, this looks like a problem with the server's network
configuration. Is there a "combo" card in the server? There are a few
autoprobing cards out there, that do not properly configure the media
type until they have received a couple of incoming data packets. If
this is a problem with your card, then try to configure it for the
media type that you have (e.g. thin wire/BNC) rather than setting it
to "autoprobe".

It could also be a problem with the ARP tables in the server, but this 
sounds less likely.

 > 2- When the client did find the serveur, 
 > it says : Loading /xx/xxx/ ... Tagged file format 
 > and the /\- signe truns very slowly.
 > This finishes 6,7 or 10 time with 
 > "Unable to load config file" and then "<Sleep>"
 > and after a moment, the load finishes correctly 
 > (many minutes after the begining of the load)
 > with the first lines of the kernel linux...
 > 
 > Here i don't understand 
 > 1- why it is slow
 > 2- Why i work sometimes and don't others.

Looks like a server problem as well. If you start bootpd and tftpd out 
of "/etc/inetd.conf", then there is a sudden peak performance
requirement on your server whenever the client boots. This might be,
what you are seeing. Check for bottlenecks (e.g. running top),
increase memory, remove other processes, watch for lost packages on
the ethernet, analyze the ethernet traffic (using tcpdump). This
should shed some light on the problem.

 > 
 > Facts :
 > 
 > Server config :
 > Pentium, last stable linux debian install this day. 
 > Ne2000.PCI
 > or 3C503 (same pb)
 > 
 > 
 > Client Config :
 > 486 Sx 33Mhz
 > Ne 2000 ISA
 > 
 > 
 > When i boot the client with his local disk
 > and run tftp to load the boot file or bootptest,
 > to see the bootp params, those are working perfectly.
 > tftp transfer time 2seconds.
 > 
 > After some times it boot whith 3.0 etherboot version.
 > I did not saw the client boot with 3.1 or 3.2.

This is strange. 3.1 and 3.2 should work just as well. Try using a
custom-made image, rather than using the pre-build ones.

 > It used to work with a 486Dx 33Mhz and a NE2000 that i don't have any more.
 > 
 > 
 > Same symthoms with the net of a friend wo has got
 > a pentium server configured with
 > a debian distribution and a pentium client,
 > and who is using ne2000 i think.
 > 
 > Thanks for any help.
 > D.G.

-- 
Markus Gutschke                         Internet: markus@infoscape.com
Infoscape, Inc                          Phone:    +1-415-537-3778
657 Mission Street, Suite 200
San Francisco, CA 94105

Disclaimer: The above message represents my personal opinion; It does
            not constitute an offical statement by Infoscape!

PGP signature