I think I have at last got this fixed. So this email is really for the benefit of the
archives.
1. Cobbler 2.6.2 has just been announced. So I've upgraded our troublesome test
environment from 2.5.0 to 2.6.2. I don't think that detail is relevant, because the
problem persisted. But at least it puts us on the latest, production quality, code.
2. Between 2.0.11 and more recent releases, the "sample.ks" file, on which our
own ".ks" files are based, had subtle changes:
$kickstart_start => $SNIPPET('kickstart_start')
$kickstart_done => $SNIPPET('kickstart_done')
(and there was an additional "%post --nochroot" clause).
Overall the snippet-ification seems to be a good thing, but the relative silence on
this transition issue caught us out. But this itself wasn't the cause of our problem.
(The snippet things would have caused us additional woes if we had used triggers, etc.)
3. With the help of Google and caffeine, I eventually tracked down that the files
"/var/lib/cobbler/config/distros.d/<foo>.json" had a null
"source_repos" attribute compared to their fully-fledged equivalents on our
production old-but-good service. This seems to be related to bugs #459 and #590. I
presume these bugs afflicted cobbler-2.5.0 (and the results stayed in place in our config
when I replaced the RPM up to 2.6.2). Is that presumption correct? When I edited the
".../distros.d/<foo>.json" file that seemed to cure the problem and the
resulting "/etc/yum.repos.d/cobbler-config.repo" looks healthy.
-- David Lee, ECMWF
----- Original Message -----
From: "David Lee" <david.lee(a)ecmwf.int>
To: "cobbler mailing list" <cobbler(a)lists.fedorahosted.org>
Sent: Friday, 27 June, 2014 11:46:13 AM
Subject: "/etc/yum.repos.d/cobbler-config.repo" (almost) empty
Production environment:
RHEL5.8 server; Cobbler 2.0.11-2 (via RPM)
Test environment:
RHEL6.4 server; Cobbler 2.5.0-172.1 (via RPM)
The two environments should be more or less the same overall, but
obviously "test" is there for trying to move forwards, then
eventually pull "production" into line with it. In both
environments we are (successfully) installing RHEL6.4 clients; the
total is about 100 machines.
"Production" is long established and works well. But I'm hitting a
problem in "test".
Installed systems in both environments include a cobbler-generated
file "/etc/yum.repos.d/cobbler-config.repo". In the "production"
environment, the file is good and contains several clauses:
[core-0]
name=core-0
baseurl=http://iii.jjj.kkk.lll/cobbler/ks_mirror/rhel64-x86_64
enabled=1
gpgcheck=0
priority=1
repeated up to "[core-5]". This is followed by a two "# error: could
not read repo source: ..." lines about two local repos, but I'm not
worried about them; they are unimportant for our environment.
But in the "test" environment those "core-<n>" clauses are
absent.
All we have are those (unimportant) "# error: ..." lines.
Are these missing "[core-<n>]" clauses a known issue? I realise that
these days even our test server "2.5.0" is out of date. So an
answer of "a known fault between 2.0 and 2.5, subsequently fixed in
2.6" would be fine! I'm just trying to get a toehold on this issue
so that I can move forwards.
Thanks.
-- David Lee