[PATCH] removed a lot of useless spaces in koan
by Christian Berendt
Hi.
I started working on the provisioning of images using cobbler (I tried
it with the latest release 2.2.1 and got a few exceptions on the client
side). Before changing code I removed a bulk of useless spaces in the
sources of koan. I attached the patch, hope it's useful to beautify the
code a little bit.. ;)
I hope the patch is in a correct form, I read
https://fedorahosted.org/cobbler/wiki/PatchProcess, but it's the first
time working with git.
Patches for the image based provisioning will follow...
Bye, Christian.
--
Christian Berendt
Linux / Unix Consultant & Developer
Mail: berendt(a)b1-systems.de
B1 Systems GmbH
Osterfeldstraße 7 / 85088 Vohburg / http://www.b1-systems.de
GF: Ralph Dehner / Unternehmenssitz: Vohburg / AG: Ingolstadt,HRB 3537
12 years, 6 months
buildiso further observations
by Simon Woolsgrove
Hi Jorgen
We found a few more issues with buildiso when moving across from 1.6 to 2.2
DNS entries
If added via the kernel options (dns=x.x.x.x,x.x.x.x) are being set as a
list on the APPEND line.
Old 1.6 entries we have used this a lot, this is a temporary fix put in
place.
@@ -349,7 +353,10 @@
if dist.breed == "suse":
append_line += " nameserver=%s" % my_dns[0]
if dist.breed == "redhat":
- append_line += " dns=%s" % ",".join(my_dns)
+ if type(my_dns) == list:
+ append_line += " dns=%s" % ",".join(my_dns)
+ else:
+ append_line += " dns=%s" % my_dns
if dist.breed in ["ubuntu","debian"]:
append_line += " netcfg/get_nameservers=%s" %
",".join(my_dns)
Management Interface
As this option never existed in 1.6 buildiso will fail as the management
field is not populated with data for any interface, this is a temporary
fix to get us round the issue in the interim. Not sure of best approach
here e.g. as we have ksdevice (redhat) set would could run a script to
set False on all systems interfaces.
@@ -267,6 +267,10 @@
mgmt_ints = []; mgmt_ints_multi = []; slave_ints = []
if len(data["interfaces"].keys()) >= 1:
for (iname, idata) in data["interfaces"].iteritems():
+ # check if management exists
+ if not "management" in data["interfaces"]:
+ idata["management"] = "False"
+
if idata["management"] == True and
idata["interface_type"] in ["master","bond","bridge"]:
# bonded/bridged management interface
mgmt_ints_multi.append(iname)
We use ksdevice a lot on our multi-nic systems, would it not make sense
if management interface is selected and ksdevice is not added to kernel
options it gets added, same could be true of other OS. The best place
for this I am not sure would be easy enough to add to buildiso but its
also a nice feature to have for koan ?
BTW the name management interface seems confusing as some folks (ok me)
will associate that we say out of band management maybe boot|build
interface would have been a better name, just a thought.
Thanks again, will keep testing.
Cheers,
Simon
12 years, 6 months
Bug with saving objects in the web GUI
by James Cammarata
I ran across this tonight, and thought I'd send an email out in case
others run into it in the future.
I noticed that trying to create an object (for example a system) was
failing when there were no other objects of that kind stored in
cobbler. I tracked the issue back to a feature we introduced in 2.2,
where an object's fields were updated only if the field value changed.
I've corrected the issue, which will be included in the next release.
For now, if you run into this issue I believe you can fix it by
manually deleting all of the files in the sessions directory
(/var/lib/cobbler/webui_sessions/ by default) since that is where the
cached field information is stored. A sign that you're running into
this issue is a log message in cobbler.log similar to this:
INFO | add_item(profile); ['']
Notice the empty string - that should be the object name. The field
info is stored in a hash, where the key is "type_name" (for example:
profile_test1), so theoretically if you add an object, then delete it,
then recreate it with the exact same name you could theoretically run
into this too, where fields are not updating like you'd expect.
Overall, I'm noticing some name space pollution in the session, so I
may work on cleaning that up in the near future.
12 years, 6 months
[PATCH] Chain load authentication plugin
by James Clendenan
Hi All,
I've attached an updated version of the authentication chain loading plugin.
It should apply cleanly to 2.2.1 now, as a few directories have been moved around.
Let me know if you run into any problems with it.
James
12 years, 6 months
A couple of things
by James Cammarata
I've written an authn module for PAM (I'll commit it soon), and it got
me thinking about the web GUI security. I'd like to propose that we
move the GUI stuff to strictly HTTPS, using some rewrite rules. This
would also get rid of the VirtualHost stuff in cobbler_web.conf, since
typically the ssl.conf defines a default vhost on 443 that takes
precedence. I did some testing, and this seems to work out pretty
well, but I wanted to get a feeling for this from everyone because it
is a fairly major move. The cobbler-web RPM would also be modified to
require mod_ssl (and mod_wsgi, which it doesn't currently).
Debian/Ubuntu packaging would need to be updated accordingly.
The cobbler_web.conf file would pretty much be reduced to the following:
RewriteEngine on
RewriteCond %{HTTPS} off
RewriteCond %{REQUEST_URI} ^/cobbler_web
RewriteRule (.*) https://%{HTTP_HOST}%{REQUEST_URI}
WSGIScriptAlias /cobbler_web /usr/share/cobbler/web/cobbler.wsgi
Pretty simple overall. The main consideration is that the config would
be created as .rpmnew, so people would have to manually fix that when
updating.
Thoughts?
12 years, 6 months
[PATCH] bugfix in buildiso
by Jörgen Maas
Hi,
There's a bug in buildiso when duplicate kopt keys are used, e.g.
duplicate console= options.
Attached patch fixes that, reported and tested by Simon Woolsgrove on
cobbler-users@
Please apply to master and stable.
--
Grtz,
Jörgen Maas
12 years, 6 months