Very Computer

I'm about to do a system wide update of recommended patches on
2.6, 7 and 8 boxes, plus possibly Y2K patches if necessary.

Before I wade in with both feet, are there any known
"repercussions" of these patch clusters?

For example, a couple of years ago i was doing something
similar to some 2.6 boxes and after the alleged "upgrade" we
found that the email had stopped working, as the recommnded
patches included an "upgrade" from sun's sendmail to a quasi-
public domain sendmail version, and now none of the original
config files were compatible, the situation being "complicated"
by the fact the upgrade merely dumped on top of the original
binary rather than renaming/moving it first, and the backout
didn't do the trick!  Subsequently I was left with a steep
learning curve in sendmail configuration files and a lot of
unplanned for work!

So - any *y traps like the above likely does anybody know
with recommended patches for 2.6, 7 & 8 (and Y2K for 2.6 ?)

Cheers

Ian

This is hard to answer, as it depends a lot on which patches
you already have installed on the systems, and if they are
the FCS or later H/W releases (ie. prepatched).

Various issues spring to mind, eg. changes in the ToolTalk authentication,
a problem with the Solaris 7 kernel patch which made the system not
boot anymore; the latest Solaris 7 kernel patch AFAIK contains a
broken rmmount binary which breaks CD access via vold.

Some problems I probably never saw because not anyone uses the same
portions of the OS. E.g. I'd never notice problems with Sun's sendmail,
as it's replaced with the one from sendmail.org right after installation.

So - you'll have to do it the painful way, read the READMEs of
all patches you want to apply. The "Special Install Instructions"
should talk about all the gotchas.

hth, mp.
--
                         Martin Paul | Systems Administrator

Liechtensteinstrasse 22, A-1090 Wien | Tel: 01 4277 38803
        http://www.par.univie.ac.at/ | Fax: 01 4277 9388

I'd say, make a NFS mount, store important sys config files there before
executing patches. This wouldnt require a whole heck of alot of space cause they
are only text files (most anyways). As for the patches, you may want to hold off
on a couple cause although they fix current problems, they introduce new ones.
We'd all like to think sun is 'stable' compared to MS (which is it), but it's
not without its bugs as well.

It's a small thing, but at least one problem I've had quite recently
was with kernel patch 106541-14 on Solaris 7, which breaks volume
management.  I don't know yet if Sun has released a fix for this.
There is an "official" workaround, but I forget what it is... check
access1.sun.com or sunsolve.sun.com.  My solution is just to do
/etc/init.d/volmgt stop  and mount cd's manually.

--
Michael Maxwell <mmaxwell at spss dot com>
SPSS Systems Specialist

                                    ^^^^^^^^^^^

Ummm, isn't it just a little bit late to be thinking about installing
Y2K patches...?  (Yes, you should install them!)

--
Rich Teer

President,
Rite Online Inc.

Voice: +1 (250) 979-1638
URL: http://www.rite-online.net

I would recommend that get a complete backup of the machines before
doing patch installation.

Yes rmmount is broken in solaris 7 kernel patch, but if you call sun
support center they will give you binary.

Usually patch cluster goes fine if you haven't install any weired
patches from the jumpstart server.

cheers
gb

Quote:> I'm about to do a system wide update of recommended patches on
> 2.6, 7 and 8 boxes, plus possibly Y2K patches if necessary.

> Before I wade in with both feet, are there any known
> "repercussions" of these patch clusters?

> For example, a couple of years ago i was doing something
> similar to some 2.6 boxes and after the alleged "upgrade" we
> found that the email had stopped working, as the recommnded
> patches included an "upgrade" from sun's sendmail to a quasi-
> public domain sendmail version, and now none of the original
> config files were compatible, the situation being "complicated"
> by the fact the upgrade merely dumped on top of the original
> binary rather than renaming/moving it first, and the backout
> didn't do the trick!  Subsequently I was left with a steep
> learning curve in sendmail configuration files and a lot of
> unplanned for work!

> So - any *y traps like the above likely does anybody know
> with recommended patches for 2.6, 7 & 8 (and Y2K for 2.6 ?)

> Cheers

> Ian

Sent via Deja.com
http://www.veryComputer.com/

ones.

well, that was rather the point of rainsing this thread to try and ascertain
which of the dodgy ones to avoid/remove from the cluster !!

Didds

If you can't annoy somebody with what you write, I think there's
little point in writing.  -- Kingsley Amis

whe I checked with the client which boxes I have to deal with are Y2K
patched
the answer was "those that were here prior to 2000"  (you can guess the next
bit of this...).

So...  the logical thing to me would be to install Y2K patches on every box
on
the basis that then they are guaranteed to be covered.  As it is the only
Y2K
cluster relevant to this patch upgrade/whatever is one for 2.6 and I only
have
a few of those to deal with and of cxourse the chances are that anything
thta's arrived since 1/1/00 had been installeed with 7 or even 8....

...  but that's why folks in case you were wondering....

;-)

Didds

If you can't annoy somebody with what you write, I think there's
little point in writing.  -- Kingsley Amis

So, has anyone got any good stories of Real (TM) Millennium bugs,
i.e. those that only manifest themselves in 2001?

Or do we have to wait until 29 February? :-)

Chris Thompson
Email: cet1 [at] cam.ac.uk