On 05/25/11 12:14 PM, Jens Elkner wrote:
On Wed, May 25, 2011 at 09:07:37AM -0700, Shawn Walker wrote:
On 05/25/11 08:11 AM, Jens Elkner wrote:
Hi,
started to test zone installs on S11X (i.e. IPS pkgs).
I noticed, that it install a pretty minimal set of packages - nice
(since we have no sparse zones anymore :()!!!
However, some often needed packages (e.g. gnu-sed, sendmail, tcsh,
ipfilter, java/runtime, java/jdk) are not in this set. So we need to
post-install it. Doing this (pkg -R /rpool/zones/$zname/root install ...)
I noticed, that it always downloads these add. pkgs from oracle (~ 85 MB),
and doesn't care, whether the cache in the global zone already has them.
This is IMHO pretty dumb and actually means, that I can't setup a zone,
when I need it, because I might be offline/on GPRS/slow DSL/foreign zone;
beside the non-sense to download the same packages several times.
What makes you believe that it was actually downloading the packages
from Oracle again?
Because it says so and takes "hours" even on fast links (i.e.> 1G):
So in other words, you haven't actually verified that it's downloading
all of the content again or what content it's downloading. Yes, that
isn't easy to do, but it's a point nonetheless.
The download progress meter is displayed regardless of whether the files
are already cached. If the files are already cached, it will validate
them and then simply move on to the next one. If any are missing from
the cache, or are no longer valid, it will download them again.
I gues you refer to the pkgs installed by zone create, not the
ones installed after zone has been created, i.e.
pkg -R /rpool/zones/$zname/root install
No, I refer to both cases. The download progress meter is *always*
displayed, even if content is already cached.
That's why the questions:
So the question: Is there a way to tell zonecfg, what add. pkgs it
should install (and thus automatically fetches it from the global zones'
cache) or alternative and probably more useful: a way to tell pkg, that
it should use /var/pkg for cache lookups, downloads instead of the
default (i.e. /rpool/zones/$zname/root/var/pkg for the example above)?
You really should be installing packages from *inside* the zone. If you
did that, it would automatically use the zone proxy and thus it would
automatically use the cache that is shared between all zones.
Well, actually we have "jumpstart" scripts for ~ 70 types of service
zones which all use pkgadd -R /rpool/zones/$zname/root -d
/net/$bla/install ... to supply the SW the zones need and to mangle the
configs etc., so that when the zones come up we _know_ they are working
as expected and can be used immediately. Never had any problems with
that approach and wish to have the same functionallity wrt. IPS!
Anyway, when you are saying "should be done from *inside*" what do you
mean with "proxy". Not sure, which version you have (here it says
55881ed54b5e), but I'm doing the add. installs right now from inside
the zone and it is the same behavior - it downloads the stuff from
pkg-orcl-1.sun.com -> coffee time.
BTW: If the proxy needs a network connection, this wouldn't work either,
since many zones are isolated, i.e. they are neither allowed to NFS mount
something, contact HTTP servers or are completely isolated, i.e. not
routed at all, no way in or out ...
It may contact that server, but you haven't verified the total amount or
type of traffic that is going to the server.
And yes, the zone proxy in the global zone may need to make network
connections.
You're making lots of assumptions about behaviour here and I'd ask you
Not sure about "assumptions". IMHO the mentioned problems are facts.
No, they're interpretations based on unverified assumptions. Please
don't be so hostile.
Note my comments are based on a build 165 or newer system. If you have
an older system, there is no zone proxy.
-Shawn
_______________________________________________
install-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/install-discuss
