From: news@ai.mit.edu (news) Newsgroups: comp.sys.sun.admin Subject: Re: booting a 386i without a pnpd. Keywords: n Message-ID: <24757@life.ai.mit.edu> Date: 3 Jun 92 13:09:23 GMT References: <3591@inca.comlab.ox.ac.uk> Organization: MIT Artificial Intelligence Lab Lines: 42 From: bruce@maypo.ai.mit.edu (Bruce Walton) Path: maypo.ai.mit.edu!bruce >I recently had to help out on a network with a mix of 386is and sparcs, one >of the 386is was the ypmaster and they wished to move the master onto one of >the sparcs. The problem after doing this was you could no longer reboot any >of the 386is, you couldn't even boot single user (even though they were >diskfull machines), because the ex-ypmaster was no longer running a pnpd (and >a few other related daemons). You couldn't start a pnpd anymore as the daemon >refused to run on a machine that wasn't the ypmaster! The only >documentation I could find with the 386is was the 4.0.2 install which says >you must have the ypmaster running before you can boot any other diskfull >machines. Why!? > >The old ypmaster would still reboot after the change, can anyone tell me >what magic file this machine has that allows it to reboot (and start the >first pnpd). Can I just do the same to the other 386is so they can boot too? >At the moment the 386i is back as ypmaster, which what they originally tried >to change! I had to do this same trick once a couple of years ago... we had an old (unreliable) 386i as YPmaster, and needed to move that function to a sparc. I soon realized it was no easy task, since the 386i's need several YP maps not used by any of the other systems. (I can't recall exactly which maps those were... and thankfully have no 386i's to access :-) but if you look at the 386i's /var/yp/Makefile, you can figure it out. You just need to copy the 386i-specific maps to the sparc, and hack the sparc's /var/yp/Makefile so that the proper maps are generated in /var/yp/`domainname`. In the end, it worked! I remember that I retained the former YPmaster as a YPslave, though I don't recall if that was a necessary part of the solution. good luck, anyway... -bruce. ------- Bruce Walton bruce@ai.mit.edu MIT artificial intelligence lab 617-253-9667 545 tech sq. cambridge, MA 02139