Hi John, John Fischer píše v čt 24. 06. 2010 v 16:54 -0700: > Octave, > > Again, if a SA is using /etc/nodename to determine the hostname via a script > then they are using an inherently broken method to determine it. The reason > being is that /etc/nodename only covers a standalone system or an IP address > that is configured locally. /etc/nodename does not cover network booted > (DHCP > or RARP/bootparams) systems. >
What is broken on using documented interface with knowledge of its limits? Best regards, Milan > Thanks, > > John > > On 06/24/10 09:33 AM, Octave Orgeron wrote: > > Hi, > > > > What would probably make sense is to maintain the config file method and > > have SMF load it. However, I'd also propose the to SMF the ability to dump > > such variables back to the standard configuration files. This could be a > > good compromise and also give SA's the ability to backup and restore the > > configuration of a server through SMF. > > > > Octave > > > > > > *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* > > Octave J. Orgeron > > Solaris Virtualization Architect and Consultant > > Web: http://unixconsole.blogspot.com > > E-Mail: unixcons...@yahoo.com > > *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* > > > > > > > > ----- Original Message ---- > > From: John Fischer<john.fisc...@oracle.com> > > To: psarc-...@sun.com > > Cc: Dave Miner<dave.mi...@oracle.com> > > Sent: Thu, June 24, 2010 10:18:07 AM > > Subject: Re: System Configuration -- nodename and defaultdomain > > [PSARC/2010/223 FastTrack timeout 06/25/2010] > > > > Still looking for a +1 on the new proposal. > > > > Thanks, > > > > John > > > > On 06/23/10 11:24 AM, John Fischer wrote: > > > >> Darren, > >> > >> Right. This case could have the data in /etc files and migrate the > >> information into SMF. > >> There are several ways to deal with System Configuration: > >> > >> 1. simply use files > >> 2. use files and migrate configuration into SMF from file > >> 3. use SMF for configuration and have files as compatibility > >> 4. use SMF only > >> 5. other (there are always other ways of doing things) > >> > >> During the design phase discussion we did not see that large of a hit by > >> using the > >> SMF only method especially since nodename is only used for standalone or > >> the IP > >> address is configured locally. Thus in certain cases /etc/nodename is not > >> a guarantee > >> that the hostname is the same as what is in the file. Since this is the > >> case scripts > >> should really be using 'uname -n' to determine the hostname. When we > >> couple the > >> hit that the customer is taking in the upgrade migration from Solaris 10 > >> it was decided > >> that now was the time to make this sort of change. > >> > >> Thanks, > >> > >> John > >> > >> On 06/23/10 10:54 AM, Darren J Moffat wrote: > >> > >>> On 23/06/2010 17:25, John Fischer wrote: > >>> > >>>> I understand that this change will break some scripts that admins > >>>> or developers have. However, as someone has already pointed out > >>>> the decision to move this direction was made in the original greenline > >>>> case. We are simply executing on one aspect of that strategy. > >>>> > >>> You are but there are is also precedence in other cases where data has > >>> been moved from a file in /etc to SMF. In some of those other cases the > >>> approach was taken to load the data in the /etc file(s) into SMF. > >>> > >>> Could this case possibly do that type of thing ? > >>> > >>> > >>>> On 06/22/10 02:29 PM, Peter Tribble wrote: > >>>> > >>>>> On Sat, Jun 19, 2010 at 9:25 PM, James > >>>>> Carlson<carls...@workingcode.com> wrote: > >>>>> > >>>>>> On 6/19/10 6:52 AM, Volker A. Brandt wrote: > >>>>>> > >>>>>>>> 3. Obsolete file /etc/nodename. This file will no longer exist in > >>>>>>>> the system > >>>>>>>> > >>>>>>> This will break 1000s of LOC of scripts. I am sure you have > >>>>>>> considered the consequences your customers will be facing, and > >>>>>>> have decided that your gain outweighs their loss... > >>>>>>> > >>>>>> Really? I can't say I've ever seen a script that reads /etc/nodename in > >>>>>> preference to using "uname -n" output, nor can I find any. Providing a > >>>>>> pointer to one would probably help focus the discussion. > >>>>>> > >>>>> I had several admin scripts that I wrote, nothing major. I would prefer > >>>>> to use `uname -n`, but after about the third time that I found a system > >>>>> thinking it was called --fqdn I started to mistrust uname -n, as it > >>>>> can be > >>>>> accidentally set to something other than the real name of the system > >>>>> by incautious admins or applications, and looked at /etc/nodename as > >>>>> a more reliable source of what the hostname was *supposed* to be. > >>>>> > >>>>> > >>>> > >>> > >> _______________________________________________ > >> opensolaris-arc mailing list > >> opensolaris-arc@opensolaris.org > >> > > _______________________________________________ > > opensolaris-arc mailing list > > opensolaris-arc@opensolaris.org > > > > > > > > > > > _______________________________________________ opensolaris-arc mailing list opensolaris-arc@opensolaris.org
