Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!wuarchive!uunet!stanford.edu!msi.umn.edu!cs.umn.edu!ianhogg
From: ianhogg@cs.umn.edu (Ian J. Hogg)
Newsgroups: comp.sys.hp
Subject: Re: HP 700 series multi-user performance?
Message-ID: <1991Apr29.211153.11046@cs.umn.edu>
Date: 29 Apr 91 21:11:53 GMT
References: <825@tiamat.fsc.com> <5589@hemuli.tik.vtt.fi> <5253@network.ucsd.edu>
Organization: University of Minnesota, Minneapolis, CSci dept.
Lines: 76

In article <5253@network.ucsd.edu> corrigan@weber.ucsd.edu (Michael J. Corrigan) writes:
>In article <5589@hemuli.tik.vtt.fi> savela@tel.vtt.fi (Markku Savela) writes:
>>In article <825@tiamat.fsc.com> jim@tiamat.fsc.com ( IT Manager) writes:
>>
>>>All the info I've seen so far seems to indicate that any of the 700's
>>>should blow away an 835, so your only trouble may be in having to
>>>pay for a multi-user license.
>>
>>  How does HP-UX count a user that must have a licence? While back I
>>someone mentioned in a article, that it would even count individual
>>X windows or at least clients as a user each. This got me a bit
>>worried and we called local HP. At least the local salesman here replied
>>that actually 2 users licence is enough for any number actual users,
>>if they come in from network connection (e.g. ethernet device is one
>                                               ^^^^^^^^^^^^^^^^^^^^^
>>user, console device another; any number of telnet sessions and x
>^^^^^ ^^^^^^^^^^^^^^^^^^^^^^^  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>>clients could be started without exceeding the user limit).
>>
>>  Was the person giving wrong info or not?
>>
>
>	I would have to say the that would be a rather generous 
>implementation of a 2-user license. In my experience it doesn't work
>that way. It seems close to correct to say that you are allowed
>2 distinct (username,hostname/tty) pairs.
>Thus if a person logs in on the console and another over a tty, then
>no network logins via rlogin or telnet are allowed.
>
>Another example, if userA logs in via telnet from hostA and userA also logs
>in from hostB via rlogin/telnet, then no tty logins (including console)
>are allowed and no further rlogins except by userA from either hostA or
>hostB. Any number of these latter are allowed.
>
>In practice on our 2-user 834's:
>"xlogin" is run by init. The console user logs in ( and stays logged
>in perpetually running as many clients as they want). Say that 
>they then also log in over the ethernet from some host.
>Then no further logins succeed over the ethernet even root, except for
>more logins by the same person from the same remote host. If it is the same
>person from a second distinct remote host, then
>"Sorry. Maximum number of users logged in."
>
>This actually seems restrictive. We had anticipated two usernames with
>any number of connections via both ttys and rlogin being allowed.
>
>I haven't figured out how it actually figures out the maximum users, though.
>There are a few situations where I sometimes seem to get 3 distinct logins.
>It would be nice to know what the actual algorithm is ( if login is looking
>at utmp entries that have certain values in certain fields, that would
>rather easy to circumvent - I don't think that is how it is done)
>
>-Mike Corrigan
>corrigan@ucsd.edu

This what you can actually do with a 2 user license (at least on our 300's &
400's) from HP-UX 6.x to the last version I used (7.something).  You are
allowed to have logins on 2 devices simultaneously.  A device is either a
tty, the console, or a pty.  All logins on pty's count as a single device. 
With this scheme we have had much more than 5 users logged in via telnet,
remsh, and whatever.

When I ran xdm on the HP's the kernel never thought that more than one user 
was logged in.  When I switched to VUE howver, I found out it's login manager
logged a login on the console.  Then the user's hpterm's logged in via pty's.
This locked out the machine for uucp dialin's because it thought 2 users were
logged in.

I wrote a program to fix the utmp entry so that console counted as one of the
pty logins.  HP sent me a patch for login to bypass this problem.  I see no
reason whatsoever for the faked login on the console.

-- 
===============================================================================
Ian Hogg						ianhogg@cs.umn.edu
                                                        (612) 225-1401