Should uname -n be reporting an FQDN?

Should uname -n be reporting an FQDN?

Post by Phillip Vand » Tue, 16 Apr 1996 04:00:00



I noticed that when I installed Solaris 2.5, both on Sparc
and x86, the nodename gets automatically set to the
machine's FQDN.

(This may have been the case for Solaris 2.4, I suppose, but
I don't know because I never allowed the hostname to be
determined automatically at installation time)

Traditionally, the nodename of the machine is only a single
name and programs are expected to get the FQDN by calling
something like:

        uname(&utsname);
        gethostbyname(utsname.nodename);

or, using BSD compatibility interfaces,

        gethostbyname(gethostname()).

I figure having the FQDN there is probably a good idea, IMHO,
especially since there is broken software which does _not_
do the above to get the FQDN, but I wonder about it because
I've never seen any other machine configured that way. Even
Sun's engineers' own machine isn't that way (in.telnetd
reports uname -n directly)

Trying 192.9.5.5 ...
Connected to playground.sun.com.
Escape character is '^]'.

UNIX(r) System V Release 4.0 (playground)

login:

When you telnet to Iodine.Mlink.NET, my telnetd tells you
its FQDN.

What are the recommended contents of /etc/nodename (which
determines uname -n, /usr/ucb/hostname, etc..)?

-Phil

 
 
 

Should uname -n be reporting an FQDN?

Post by Greg Linda » Wed, 17 Apr 1996 04:00:00




>I figure having the FQDN there is probably a good idea, IMHO,
>especially since there is broken software which does _not_
>do the above to get the FQDN

like statd. If you run a site with multiple domains which lock between
domains, you need hostnames which are fully qualified.

SGI, by the way, has fixed statd.

-- g

 
 
 

Should uname -n be reporting an FQDN?

Post by bch.. » Wed, 17 Apr 1996 04:00:00


: I noticed that when I installed Solaris 2.5, both on Sparc
: and x86, the nodename gets automatically set to the
: machine's FQDN.
:
: (This may have been the case for Solaris 2.4, I suppose, but
: I don't know because I never allowed the hostname to be
: determined automatically at installation time)
:
: Traditionally, the nodename of the machine is only a single
: name and programs are expected to get the FQDN by calling
: something like:
:
:       uname(&utsname);
:       gethostbyname(utsname.nodename);
:
: or, using BSD compatibility interfaces,
:
:       gethostbyname(gethostname()).
:
: I figure having the FQDN there is probably a good idea, IMHO,
: especially since there is broken software which does _not_
: do the above to get the FQDN, but I wonder about it because
: I've never seen any other machine configured that way. Even
: Sun's engineers' own machine isn't that way (in.telnetd
: reports uname -n directly)
:
: Trying 192.9.5.5 ...
: Connected to playground.sun.com.
: Escape character is '^]'.
:
:
: UNIX(r) System V Release 4.0 (playground)
:
: login:
:
: When you telnet to Iodine.Mlink.NET, my telnetd tells you
: its FQDN.
:
: What are the recommended contents of /etc/nodename (which
: determines uname -n, /usr/ucb/hostname, etc..)?
:
: -Phil

        I recommend using FQDN wherever possible. I have
totall sworn off short names. Use it in both the /etc/hostname.*
files and /etc/nodename.

   ____________________________________________________________________

  | Systems Programmer              University of California, Davis    |
  | Graduate School of Management & GIS Center                         |

 
 
 

Should uname -n be reporting an FQDN?

Post by Tom Sorens » Fri, 19 Apr 1996 04:00:00



>I noticed that when I installed Solaris 2.5, both on Sparc
>and x86, the nodename gets automatically set to the
>machine's FQDN.

>(This may have been the case for Solaris 2.4, I suppose, but

Well, it's not on my system. My university account finally expired, so I
can't check the multiple Solaris 2.4 systems there (which, in theory, would
be correctly setup given how much time Sun spent trying to make the things
work as advertised. Solaris just doesn't handle several hundred interactive
users well)


dogbert

Quote:>What are the recommended contents of /etc/nodename (which
>determines uname -n, /usr/ucb/hostname, etc..)?


dogbert

A FreeBSD system I have access to reports the FQDN. A SunOS 4.1.3 system
does not. Your guess as to which is right.

--

If I managed to represent TI in this post, then I'm    | Responsibility | () |
probably far more surprised than TI is.                | Starts at Home | /\ |

 
 
 

Should uname -n be reporting an FQDN?

Post by ron na » Fri, 19 Apr 1996 04:00:00



: >I noticed that when I installed Solaris 2.5, both on Sparc
: >and x86, the nodename gets automatically set to the
: >machine's FQDN.
: >
: >(This may have been the case for Solaris 2.4, I suppose, but

Not on our 2.4 systems:
gondor.nash:~> uname -a
SunOS gondor 5.4 Generic_Patch sun4c sparc

: Solaris just doesn't handle several hundred interactive users well)

Umm - I think it does if you configure it right!  Here are current uptime
loads from our SS1000E running Solaris 2.5 (rohan.sdsu.edu):

  2:00pm  up 7 day(s),  6:29,  165 users,  load average: 3.23, 3.70, 3.60
  2:10pm  up 7 day(s),  6:39,  174 users,  load average: 2.84, 3.11, 3.33
  2:20pm  up 7 day(s),  6:49,  175 users,  load average: 4.40, 4.28, 3.81
  2:30pm  up 7 day(s),  6:59,  169 users,  load average: 4.20, 4.37, 4.07
  2:40pm  up 7 day(s),  7:09,  174 users,  load average: 4.41, 4.52, 4.29

--

  ,;( )__, )~\|  |
 ;; //   '--;    | Gin-N-Tonic   endurance horse
 '  ;\    |      | Luv on Fire   trusty trail horse

 
 
 

Should uname -n be reporting an FQDN?

Post by Tom Sorens » Tue, 23 Apr 1996 04:00:00




>: Solaris just doesn't handle several hundred interactive users well)

>Umm - I think it does if you configure it right!  Here are current uptime
>loads from our SS1000E running Solaris 2.5 (rohan.sdsu.edu):

>  2:00pm  up 7 day(s),  6:29,  165 users,  load average: 3.23, 3.70, 3.60

That's right about the limit. Try adding another one or two hundred. Solaris
starts * badly. Solaris 2.3 spent 80% of it's time in overhead at
around 180 users. It would crash soon thereafter.

The system in question was also doing campus wide NFS serving, serving
news to local users from another internal machine (local databases, updates
were coming from another machine inside the local subnet, which was fed
by another machine, which talked to the rest of the world), and being used
for just about everything you'd expect in a technical university environment
(including morons who would run CPU crunchers on it instead of the
supercomputers available).

Sun wound up sending multiple engineers out at their cost to fix the problem.
They couldn't. At this point the system has devolved into several large
systems, including a 12 CPU SS2000 w/ 1G RAM (the original box was this w/
6 CPUs and 512M RAM), 2 SS1000s w/ 6 CPU & 512M RAM, and 2-4 SS20's. One
server is now dedicated to NFS, another to news, and three run as interactive
systems. Between them they're handling approximitely 300-400 users with no
more than a crash every month or so. Original configuration (and the first
few quick fix configs) crashed hourly, daily, or weekly depending upon luck.

Much of the code in Solaris 2.4 and 2.5 that improved multiuser performance
came about as a part of the above. Many sites use Solaris for multiuser, but
not on the scale of 300-500 users on a single box with NFS sharing and news
service.

--

If I managed to represent TI in this post, then I'm    | Responsibility | () |
probably far more surprised than TI is.                | Starts at Home | /\ |

 
 
 

Should uname -n be reporting an FQDN?

Post by Frank Pete » Wed, 24 Apr 1996 04:00:00





> >: Solaris just doesn't handle several hundred interactive users well)

> >Umm - I think it does if you configure it right!  Here are current uptime
> >loads from our SS1000E running Solaris 2.5 (rohan.sdsu.edu):

> >  2:00pm  up 7 day(s),  6:29,  165 users,  load average: 3.23, 3.70, 3.60

> That's right about the limit. Try adding another one or two hundred. Solaris
> starts * badly.

Actually, there is a patch to Solaris 2.4 for this and Solaris 2.5 has
the patch bundled.  With this configuration you can easily get three or
four hundred users on a suitably configured system (probably more, that's
just the most I've seen).

Also note that the same problem actually exists in SunOS, its just very
rare to see a SunOS system try to support a couple of hundred or more
users.  It probably also exists in other traditionally-desktop systems.
--
Frank Peters - UNIX Systems Group Leader - Mississippi State University

                <URL:http://www.veryComputer.com/~fwp/>