garvey@netcom.UUCP (Joseph Garvey) (07/23/89)
I have a network of HP 9000/3xx machines, well actually a cluster in the
HP parlance. The server is named "zeus". I'm in the process of hooking us
up to Usenet, and noticed that someone is already using the name zeus. We
have a significant investment in local software customized to particular
workstations. We also have quite a few computer semi-(il)literates that would
be total confused for months if we changed the name of the server from zeus.
So I'm stuck... my lan wants me to be zeus, while usenet wants me to be cmic.
I use the myname option in the uucp/Permissions file to identify myself as
"cmic" to my feed, and that works, that is information will correctly
transfer. However, after the connection is closed and uux runs to post mail,
it sees an address of cmic!username and gets confused (it thinks the name
should be zeus... guess it doesn't look in the Permissions file).
How do I fool the mail system into finishing the job uucp started. I've looked
at the sendmail.cf file hp supplies, and can actually fool it into delivering
inbound mail... but the then outbound mail gets screwed up.
Is what I'm trying to do total crazy? Am I in for real trouble when I try to
add news? Should I just bite the bullet and convert the system name to cmic,
and find/fix all the local software (it'll take months to weed everything
out). Not to mention HP doesn't support changing the name of a node in a cluster
short of deleting everything and starting again. (I've done it once early on,
it can be done... it takes quite a while...). And of course the users will
howl (bawl, whine)... what the heck... they're only a test load any way. :-)
Would it be totally out of the question to call our system zeus on Usenet? From
what I can tell about addressing the path is always explicitly supplied...
and that would eliminate confusion. Or would this be considered immoral? :-)
Thanx.
Joe Garvey
California Microwave
Sunnyvale Ca
408-720-6439 (9-6, M-F), let it ring
{backbone...}!amdahl!netcom!garvey
{uunet...}!apple!netcom!garveyaccu@hpuepta.UUCP (Accugraph Corp) (07/29/89)
I had a similar dilemma when I hooked up my system via uucp to our local HP office. What I ended up doing was making my previous hostname (tigger), an alias in /etc/hosts: BEFORE: 192.1.2.11 tigger pw phil AFTER: 192.1.2.11 acgrph tigger pw phil Note that the official hostname changed from tigger to acgrph, but virtually all net services will still work with tigger. gethostbyname(3N) looks at the aliases as well as the official name, hence ARPA, Berkeley and NFS still work fine. E.g.: % ftp acgrph % rlogin tigger % mount pw:/users/pw /users/pw There may be a way to reconfigure sendmail/uucp, but I could not figure it out. Hope this helps. pw uucp: hpfcse!hpuepta!acgrph!pw
puglia@cunixc.cc.columbia.edu (Paul Puglia) (07/31/89)
I believe there is a macro in the sendmail.cf file which allows you to set what host name sendmail uses when it mails things to be. Usually, you just set this to the hostname with one of the built in macros, but it is possible to manually set this to some other name if you want. I did this with a machine at a former place of empoly. I had the machine know by one name internally and known by another on USENET. I found the macro by looking through the comments in the sendmail.cf file. Paul Puglia Dept of Civil Engineering Columbia University Internet: puglia@cucevx.civil.columbia.edu Bitnet: PUGLIA@CUCEVX