dmr@csli.Stanford.EDU (Daniel M. Rosenberg) (05/07/89)
I had just gotten 0.8 WriteNow to print on my LaserWriter (thanks for the help on that!) and was seriously into hacking the hell out of my system into workability. Mail off of our next box never made it further than the Mailer-Daemon; it complained something about how mailer ether didn't know how to get to mailhost, and I couldn't figure out how to tell it what mailhost was. So, I thought, why not just get yellow pages running on this sucker, so as to make it have the same idea about centralized network services as all the other machines in our area. Simple enough, with ypinit -s, but when I /etc/reboot'ed to implement this, our machine hung -- it did a ypbind, then stuck at the inetd spot. So, now we have this NeXT box that won't boot. Is there any well to interrupt the rc.local (or whatever) script and tell it to just move on, so I can fix the error? Or is the only thing I can do to take out the disk, bring it to another NeXT box, and try to repair it there even though I don't quite know what I did wrong? Or do I have to end up waiting for 0.9 to arrive? (We didn't send in the registration card yet. Dumb, dumb, dumb.) Glad for any suggestions -- I remain, -- # Daniel M. Rosenberg // Stanford CSLI // Opinions are my own only. # dmr@csli.stanford.edu // decwrl!csli!dmr // dmr%csli@stanford.bitnet
jgreely@previous.cis.ohio-state.edu (J Greely) (05/08/89)
In article <8853@csli.Stanford.EDU> dmr@csli.stanford.edu (Daniel M. Rosenberg) writes: >I had just gotten 0.8 WriteNow to print on my LaserWriter Hmmm... I must have missed that one. I'd ask you to clarify how you got it working, but I'm supposed to have 0.9 tomorrow, so I won't. >Mail off of our next box never made it further than the Mailer-Daemon; >it complained something about how mailer ether didn't know how to get >to mailhost, and I couldn't figure out how to tell it what mailhost >was. You never changed the default sendmail.cf, which thinks there is a much brighter machine hiding out on the network, under the name of "mailhost". The correct solution is to either edit (twisted path) /bootdisk/Administration/Private/conf/sendmail.cf, changing mailhost to something sensible, or borrow a better .cf from a local sendmail wizard. (Yes, 0.8 requires a fair amount of Unix expertise, as well as occasional screams/giggles about the file system arrangement. This, at least, changes in 0.9) I went the wizard route, since we have some experienced sendmail hackers here (they're the ones with the bloodshot eyes and maniacal grins). >So, I thought, why not just get yellow pages running on this >sucker, so as to make it have the same idea about centralized network >services as all the other machines in our area. One of those nice, simple things. In a perfect world, this means setting domainname, making sure the other machines know about you properly, and rebooting. It took me about half an hour here, most of it spent waiting on yppush. >Simple enough, with ypinit -s, but when I /etc/reboot'ed to implement ^^^^^^^^^ Icky! You turned yourself into a slave server. Did the master server know this, did you set up the directory to hold the maps, and did you use ypxfr to copy over current maps? If not, the machine thinks it's a YP server, but has no database. This *will* cause it to hang. >this, our machine hung -- it did a ypbind, then stuck at the inetd >spot. See? :-) Seriously, you probably don't want to be a slave server. For one thing, other machines may become clients of you, which may lower your performance enough to destroy the small win of having the databases local. Yp servers (and NFS, news, mail, and pizza servers) should be fast and ugly, and used for little else. It frees the fun machines for fun things. >So, now we have this NeXT box that won't boot. Is there any well to interrupt >the rc.local (or whatever) script and tell it to just move on, so I can >fix the error? (aside: ^C-ing most commands in rc* will kill them and move on, but the machine might not come up terribly sane. In fact, it may just sit there with a silly grin on it's console) Boot single-user. Translation: b<device> -s, where <device> is either od or sd, depending on you're booting from an Optical or a SCSI disk. You now have a very minimal environment, and can work slowly on the console. The console is stupid, so just undo enough to bring the machine up standalone (disconnect the ethernet), then make further changes from a Terminal window (as root, natch). Following that, run-don't-walk and pick up a book on Unix system administration (ask around for local recommendations). It will add immensely to your ability to cope with common problems. Beware that many books assume System V, which confuses people handily (I think there's a very nice Nutshell book on BSD, with tips of all sorts, but don't quote me). >Or is the only thing I can do to take out the disk, >bring it to another NeXT box, and try to repair it there even though >I don't quite know what I did wrong? Needless to say, no. The first, best, thing to do is locate a local expert, send them gifts of Coke and Twinkies, and then bug the h*ll out of them when you get confused. Sufficiently large bribes will cover almost any problems, except on weekends (you may have to resort to pizza, or, depending on the environment, sprouts and tofu). >Or do I have to end up waiting for 0.9 to arrive? (We didn't send in >the registration card yet. Dumb, dumb, dumb.) <grin> One way to see if you're scheduled to get it anyway is to send email to tcarlin@next.com, who is, I believe, co-ordinating such things. At least, I have mail in my office claiming that. "Give him some orange juice and a sugar cookie, ... maybe a tetanus shot. He'll be fine." -=- J Greely (jgreely@cis.ohio-state.edu; osu-cis!jgreely)
eht@cs.cmu.edu (Eric Thayer) (05/08/89)
About the sendmail stuff:
The easiest thing to do, of course, is to find a Cube with a working
/usr/lib/sendmail.cf configuration file. If one doesn't exist, what
we had to do was to change the lines:
DRmailhost
CRmailhost
to
DRcs.cmu.edu
CRcs.cmu.edu
There were some weirdnesses in the address rewrite rules that we
had to remove, but this only affected incoming mail. You can
test whether the address rewrite rules are working 'correctly'
by:
telnet <yourNeXT> smtp
and use:
vrfy you@yourmach.stanford.edu
vrfy you@yourmach
vrfy you
quit
In those cases, you should get back:
250 Your Full Name <you@yourmach.stanford.edu>
.
.
.
Have you tried booting single user and fixing up your rc script?
(type 'b -s' at the rom monitor prompt)
Hope this helps,
...eric
p.s. Where did all the BSD documentation go for 0.9? I tried to find
the Sendmail documentation and to my dissapointment all I found were
man pages refering me to the stuff I wanted.gore@eecs.nwu.edu (Jacob Gore) (05/08/89)
If you are just using /etc/hosts, or, God help you, Yellow Pages, you
should be able to just add "mailhost" as an alias to some real machine.
Jacob Gore Gore@EECS.NWU.Edu
Northwestern Univ., EECS Dept. {oddjob,chinet,att}!nucsrl!gore