Adam Rosenwald wrote: > A FAQ that I have brought up on numerous occasions on IRC involves > 'getting rid of bad cobbler metavariables'. > > If I were to set a wrong ks_meta variable NAME, how do I get rid of > the NAME without taking out ks_meta entirely. > I suspect this is not well documented, though it's been present for a while: cobbler system edit --name=foo --ksmeta="~NAME" --in-place replacing "NAME" with the key you want to delete. (Or, if this is too cryptic to remember, you can also use the web app)

Hello all! Have a little problem here :( when I run 'cobbler profile dumpvars f12-x86_64' the following error occurs: Traceback (most recent call last): File "/usr/bin/cobbler", line 35, in <module> sys.exit(app.main()) File "/usr/lib/python2.5/site-packages/cobbler/cli.py", line 501, in main rc = cli.run(sys.argv) File "/usr/lib/python2.5/site-packages/cobbler/cli.py", line 188, in run self.object_command(object_type, object_action) File "/usr/lib/python2.5/site-packages/cobbler/cli.py", line 264, in object_command data = self.remote.get_blended_data(options.name,"") File "/usr/lib/python2.5/xmlrpclib.py", line 1150, in __call__ return self.__send(self.__name, args) File "/usr/lib/python2.5/xmlrpclib.py", line 1434, in __request allow_none=self.__allow_none) File "/usr/lib/python2.5/xmlrpclib.py", line 1083, in dumps data = m.dumps(params) File "/usr/lib/python2.5/xmlrpclib.py", line 623, in dumps dump(v, write) File "/usr/lib/python2.5/xmlrpclib.py", line 638, in __dump f(self, value, write) File "/usr/lib/python2.5/xmlrpclib.py", line 642, in dump_nil raise TypeError, "cannot marshal None unless allow_none is enabled" TypeError: cannot marshal None unless allow_none is enabled The only 'None' in the profile is for Repos when viewed via cobber_web, command line reports Name : rescue-f9-i386 Comment : DHCP Tag : default Distribution : f9-i386 Enable PXE Menu? : True Kernel Options : {'rescue': '~'} Kernel Options (Post Install) : {} Kickstart : /var/lib/cobbler/kickstarts/pxerescue.ks Kickstart Metadata : {} Management Classes : [] Name Servers : [] Name Servers Search Path : [] Owners : ['admin'] Red Hat Management Key : <<inherit>> Red Hat Management Server : <<inherit>> Repos : [] Server Override : <<inherit>> Template Files : {} Virt Auto Boot : 1 Virt Bridge : br9 Virt CPUs : 1 Virt File Size(GB) : 10 Virt Path : Virt RAM (MB) : 512 Virt Type : qemu -Ron _______________________________________________ cobbler mailing list cobbler(a)lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/cobbler

Adam Rosenwald wrote: > A FAQ that I have brought up on numerous occasions on IRC involves > 'getting rid of bad cobbler metavariables'. > > If I were to set a wrong ks_meta variable NAME, how do I get rid of > the NAME without taking out ks_meta entirely. > I suspect this is not well documented, though it's been present for a while: cobbler system edit --name=foo --ksmeta="~NAME" --in-place replacing "NAME" with the key you want to delete. (Or, if this is too cryptic to remember, you can also use the web app) > `cobbler system edit --in-place --name=NAME --ksmeta='*!art2*'` > > Supposedly this syntax will, upon deserialization of the JSON, clear > the ks_meta namespace of *art2*; however, upon issuing a report on a > given entity, the flawed variable will still appear along with the > negated variable. That's not about deserialization, that's why... it's about supression of evaluation. What it means is that when evaluating the object chain the variable will be discarded. For instance, if the distro sets kernel option "a=2", and the profile sets "!a", there will be no value a in the evaluated kernel options when installing based on that profile. A better usage example if if globally in settings the kernel arguments included "text" and someone wanted to always do a graphical install for a particular profile: cobbler profile edit --name=foo --kopts="!text"

--Michael

Yes, ! suppresses the meta variable from being used by Cheetah, ~ deletes it from the list. Be sure to use --in-place when doing this, otherwise you will erase all meta variables. The ~ also works in kopts, kopts-post, etc.

Christian Horn wrote: > On Sun, Jun 28, 2009 at 03:24:01PM -0500, James Cammarata wrote: > >> Yes, ! suppresses the meta variable from being used by Cheetah, ~ >> deletes >> it from the list. Be sure to use --in-place when doing this, otherwise >> you >> will erase all meta variables. The ~ also works in kopts, kopts-post, >> etc. >> > > Any idea on why deletion doesnt work here? > > # cobbler system edit --name=chhorn_tester3 --ksmeta='aaa bbb ccc' > > # cobbler system report --name=chhorn_tester3|grep meta > ks metadata : {'aaa': None, 'bbb': None, 'ccc': None} > > # cobbler system edit --name=chhorn_tester3 --in-place --ksmeta='~aaa' > > # cobbler system report --name=chhorn_tester3|grep meta > ks metadata : {'aaa': None, 'bbb': None, 'ccc': None, '~aaa': > None} > No, I don't... Works for me (though I'm using HEAD):

James Cammarata wrote: > On Tue, 30 Jun 2009 10:24:03 -0400, Michael DeHaan <mdehaan(a)redhat.com&gt; > wrote: > >> Christian Horn wrote: >> >>> Any idea on why deletion doesnt work here? >>> >>> # cobbler system edit --name=chhorn_tester3 --in-place --ksmeta='~aaa' >>> >> No, I don't... Works for me (though I'm using HEAD): > > Did we backport this into release16? I don't think we did (and a quick > look at the code seems to confirm this). So it's not broken, this was a > new feature that was added to devel only a while back. It just seems like > it was added to the main branch a long time ago because we've been working > on so many major changes for 2.0. Most likely, in which case, look for this in the next major release!

> > Thanks for clearing up the distinction between 'suppression of > evaluation' vs 'deserialization'. I wasn't quite sure how the '~' or > '!' prefix applied to kernel options/kickstart variables when being > parsed by anaconda. If the variables are not thrown out during > deserialization, then the '!' or '~' prefixes /do/ in fact have > meaning for the anaconda parser (unless cobbler does additional > processing before the variables are passed to anaconda's parser). Is > there a distinction between the '~' or '!' prefixes? Try to avoid saying "deserialization" here if you can. This isn't the development list, and I don't want folks to feel like they have to understand what that means :) Actually you mean "serialization", that is, when the files are saved. (Further, we also don't want folks to have to know the JSON files exist either... the on-disk format isn't terribly important and the idea is folks do everything via Cobbler).

(A) "~option" -- Delete this immediately from the internal saved state... In other words, I put a value here and now I want to delete it and pretend I never put that value there. Also, I don't want to use the web app to do it. This would remove the value from the internel saved state. (B) "!option" -- I know something higher up is going to define option, and I want to pretend as if they didn't set it, so document the absense of this flag there and make it not be applied. (This is understandably complex) Both syntaxes are odd mainly because there wasn't a cleaner way to do the syntax without introducing another flag, and we /really/ didn't want to do that. > > My original question, however, relates to whether and how removing the > kopts/ksmeta variables can be achieved, such that they are completely > gone -- obliterated from serialized JSON files. No trace of them > whatsoever in cobbler hashes. Is there currently a way to do this via > the CLI? If not, do you have any objection to my submitting patches > on cobbler-devel? Yes with "~option".