Checking for existing IP addresses can sometimes be a bad thing.
by Dave Hatton
I wanted to discuss this before pushing onwards to try to fix my problem.
To set the scene; I use cobbler to build Oracle RAC clusters which use a
private network for cluster communications on either separate hardware
or a VLAN. (I'm pretty sue most other clusters do the same.) Generally,
I use 10.0.0.n for the addresses in this private network.
Now the problem is when I add these interfaces for a my 2nd cluster I
get a error message complaining that the IP is (quite rightly) already
in use.
It makes sense to me to reuse the same 10. address range for Oracle
cluster interconnect because this simplifies installation/maintenance
somewhat. Equally I can see why we want to validate IP addresses for
most other purposes.
So, how about some form of override ?
Dave Hatton
15 years, 4 months
Re: Checking for existing IP addresses can sometimes be a bad thing.
by James Cape
Being able to override is also useful for anycast services (e.g. DNS, TFTP).
-JC
------Original Message------
From: Dave Hatton
Sender: cobbler-bounces(a)lists.fedorahosted.org
To: cobbler mailing list
ReplyTo: lists(a)davehatton.it
ReplyTo: cobbler mailing list
Subject: Checking for existing IP addresses can sometimes be a bad thing.
Sent: Jan 21, 2009 6:43 AM
I wanted to discuss this before pushing onwards to try to fix my problem.
To set the scene; I use cobbler to build Oracle RAC clusters which use a
private network for cluster communications on either separate hardware
or a VLAN. (I'm pretty sue most other clusters do the same.) Generally,
I use 10.0.0.n for the addresses in this private network.
Now the problem is when I add these interfaces for a my 2nd cluster I
get a error message complaining that the IP is (quite rightly) already
in use.
It makes sense to me to reuse the same 10. address range for Oracle
cluster interconnect because this simplifies installation/maintenance
somewhat. Equally I can see why we want to validate IP addresses for
most other purposes.
So, how about some form of override ?
Dave Hatton
_______________________________________________
cobbler mailing list
cobbler(a)lists.fedorahosted.org
https://fedorahosted.org/mailman/listinfo/cobbler
Sent from my Verizon Wireless BlackBerry
15 years, 4 months
cobbler reposync
by Scott McClanahan
I've read the man page and came across this:
"... This is probably a command you would want to put on a crontab,
though the frequency of that crontab and where the output goes is left
up to the systems administrator..."
it's been left up to the sysadmin to make sense of the output and I will
unless there is something else out there that already does this.
Initially, I'd like to see what changed in each repo after the reposync.
A list of packages sync'd would be great. Is there any reporting like
this already?
Also, does createrepo get run regardless if anything within the repo
actually changed? Thanks.
15 years, 4 months
Checking for existing IP addresses can sometimes be a bad thing.
by Dave Hatton
I wanted to discuss this before pushing onwards to try to fix my problem.
To set the scene; I use cobbler to build Oracle RAC clusters which use a
private network for cluster communications on either separate hardware
or a VLAN. (I'm pretty sue most other clusters do the same.) Generally,
I use 10.0.0.n for the addresses in this private network.
Now the problem is when I add these interfaces for a my 2nd cluster I
get a error message complaining that the IP is (quite rightly) already
in use.
It makes sense to me to reuse the same 10. address range for Oracle
cluster interconnect because this simplifies installation/maintenance
somewhat. Equally I can see why we want to validate IP addresses for
most other purposes.
So, how about some form of override ?
Dave Hatton
15 years, 4 months
Koan seems unable to manage my LVM storage
by Fabien Dupont
Hi.
I'm quite new to Cobbler but like the abilities of system management
it provides.
I have configured my Cobbler installation to manage RHN repos, DHCP
and DNS records.
Now I try to use Koan to deploy virtual machines. My server
configuration uses LVM for VMs system disk storage and iSCSI for data
storage.
My VG is named 'vmsysvg' and its size is 9,32 GB. And Koan can't
handle this number. Here are the logs :
# koan --server=cobbler.virtual.local --virt --system=xenguest01.virtual.local
- reading URL: http://10.0.0.1/cblr/svc/op/ks/system/xenguest01.virtual.local
install_tree: $tree
libvirtd (pid 3677) en cours d'exécution...
downloading initrd initrd.img to /var/lib/xen/initrd.img
url=http://cobbler.virtual.local/cobbler/images/rhel-5.2-server-x86_64-xen/initrd.img
- reading URL: http://cobbler.virtual.local/cobbler/images/rhel-5.2-server-x86_64-xen/in...
downloading kernel vmlinuz to /var/lib/xen/vmlinuz
url=http://cobbler.virtual.local/cobbler/images/rhel-5.2-server-x86_64-xen/vmlinuz
- reading URL: http://cobbler.virtual.local/cobbler/images/rhel-5.2-server-x86_64-xen/vm...
vgs -o vg_name
File descriptor 3 left open
File descriptor 4 left open
VG
datavg
iscsivg
rootvg
vmsysvg
vgs --noheadings -o vg_free --units g vmsysvg
File descriptor 3 left open
File descriptor 4 left open
(9,32)
exceptions.ValueError
invalid literal for float(): 9,32
File "/usr/lib/python2.4/site-packages/koan/app.py", line 191, in main
k.run()
File "/usr/lib/python2.4/site-packages/koan/app.py", line 408, in run
self.virt()
File "/usr/lib/python2.4/site-packages/koan/app.py", line 683, in virt
return self.net_install(after_download)
File "/usr/lib/python2.4/site-packages/koan/app.py", line 602, in net_install
after_download(self, profile_data)
File "/usr/lib/python2.4/site-packages/koan/app.py", line 681, in
after_download
self.virt_net_install(profile_data)
File "/usr/lib/python2.4/site-packages/koan/app.py", line 1089, in
virt_net_install
path_list = self.calc_virt_path(pd, virtname)
File "/usr/lib/python2.4/site-packages/koan/app.py", line 1360, in
calc_virt_path
path = self.calc_virt_path2(pd,name,offset=count,location=x,sizes=virt_sizes)
File "/usr/lib/python2.4/site-packages/koan/app.py", line 1412, in
calc_virt_path2
freespace = int(float(freespace_str))
Maybe the coma sign is not what's awaited by python.
Hope you can help me...
--
Fabien Dupont
System Engineer
15 years, 4 months
SELinux denies tftpd access to vmlinuz-PAE
by Tim Largy
I'm running Fedora 9, using cobbler 1.4.1 that yum retrieved for me
from the Fedora updates repository. SELinux is denying tftpd access to
vmlinuz-PAE. This is in my /var/log/messages many times:
setroubleshoot: SELinux is preventing in.tftpd (tftpd_t)
"read" to ./vmlinuz-PAE (httpd_sys_content_t)
By the way, SELinux also prevents dhcpd and tffpd from writing to
/var/log/cobbler/cobbler.log. I'm not sure I'm in the mood to learn
how to write SELinux policy so I'm going to turn off SELinux
enforcement and try again.
Tim
15 years, 4 months
[devel and somewhat rambling] Calling all you SQL geeks; opinions wanted
by Michael DeHaan
So, I'm doing the SQL thing full-on after all.
I think ideally I'd like either a way for folks using serializer_catalog
now to upgrade to serializer_sqlite through a seperate tool/step, or
otherwise make serializer_sqlite know about the older serializer so it
can upgrade things automatically when folks switch over (like
serializer_catalog did for serializer_yaml -- this saves me writing a
migration tool).
I have checked in a very wrongish initial schema to "db/*.sql" in the
cobbler tree.
We are /not/ using sqlalchemy. Why? (A) Databases are a pain to set
up. (B) For any imaginable number of systems sqlite is going to be all
we'll need. (C) the level of complexity adding an ORM to the project
brings isn't worth it when we also need to support the existing text
based serializers (YAML, which I still like).
So, the question is thus... to normalize or not to normalize?
We could do a pretty straight translation of the YAML, storing each in
it's own field in the SQL. This would require some python evals but
would most definitely work. The problem is it limits searchability in
ways that are not good.
We at minimum need to make at least one table for something that is
/not/ a cobbler object -- the interfaces that are part of the system
objects, since they are so complicated and we frequently need to search
on mac/ip.
Once all of this is done and we have this serializer, there are two
fundamental code changes that are going to happen. Anywhere where we
currently loop through objects is going to need to be changed to use a
different iterator, to loop instead through object /names/ in many
cases, only creating an object from what's stored in the database when
they need to be used. Also, we need to make sure that the find() code
in the item/collections code nows that when using the database
serializer, it doesn't have to iterate -- it needs to know about how to
generate sql statements at this point.
How complicatd this gets depends a bit on how normalized things get.
I'm presently against trying to make a seperate table for each
hash/array and so forth.
Another concern of mine is that adding new fields to cobbler remains
simple. As we move forward, adding new fields to item_*.py will also
require creating a schema change in the "db/" directory, for those using
the SQL serializers, and it will require bumping the monotonically
increasing SQL version.
Apologies if this is hard to share in text form, or if email is bad for
questions, but I did want to share it -- and I can try to elaborate more
on what I was thinking if needed. I think it's better to err on "I
don't understand what you are talking about" than to have not shared it
and then folks might be confused with the result.
So, for those hacking on cobbler (who will be impacted a minor amount by
whatever we do here), comments welcome.
I'm sure I can make something workable in any case, though I want to
keep things simple to use and also keep existing configurations
working. We've done that for a long time in Cobbler land, and I'm not
about to break forward compatibility to switch to a pure-database
system, because I think that would do bad things to the userbase.
We could /possibly/ explore a migration tool, but I think there is
beneficial logic in the way cobbler self-upgrades itself today. (for
instance, when cobbler's schema changed, it just worked, and it didn't
matter how it was being saved at the time -- this should still be true
moving forward -- even if there is some extra work to be done on the
side to keep the SQL parts in sync with this).
--Michael
15 years, 4 months
Cobbler ticket #344 - problem with debian preseeds
by Jayson Barr
Hello all,
I am having a problem getting preseed files to work with cobbler and
was hoping for some help.
I filed a ticket here:
https://fedorahosted.org/cobbler/ticket/344
Someone responded that I should bring this up in the mailing list, so here I go.
Here I will basically past everything from the ticket with a little at
the end about some recent developments:
I am running cobbler 1.3.4 on FC8. I am successfully using it to
automate Fedora Core 8 and 10 installations with my customized
kickstarts. Truly, cobbler is fantastic, but I am having a lot of
problems using it to preseed custom ubuntu installations.
The short story: I can't get cobbler to use my preseed file.
The long story: First off, using import results in the creation of
various directories in /var/lib/cobbler and /var/www/cobbler, which is
expected, but does /not/ produce a "repo" or "distro" as far as
cobbler is concerned.
After running a multitude of "cobbler sync" and "cobbler reposync"
commands, restarting the cobbler daemon, more syncing, no matter what
I do: when I ran "cobbler list" there were no repo, distro, or profile
entries of ubuntu. I specifically named them all Ubuntu_8.04, and only
fedora repos, distros, and profiles came up.
So, I looked at the DebianDeployment page some more and basically did
the old suggested steps of importing a full install disc as well as a
netboot image. Neither import worked, so I did "cobbler repo add" and
"cobbler distro add" with them and now they appear in "cobbler list"
output. I used "cobbler edit" to make the normal Ubuntu distro use the
kernel and initrd from the netboot image.
Now, I have successfully gotten the Ubuntu installer to load over my
network with PXE, but it is not sending the preseed file or I have no
idea how to tell the computer being installed to pick it up.
When I created the profile for this installation method, I used
--kickstart=/path/to/preseedfile.cfg
I also used breed=ubuntu on the distro.
Anyone have any clues or suggestions?
Thank you very much for your time.
Addendum:
I was able to just put up my preseed file elsewhere on my webserver
and then added this kernel option to the cobbler profile:
cobbler profile edit --name=Ubuntu_8.04 --kopts="lang=en_US
preseed/url=http://myserver/ubuntu_ks/kickstart_Ubuntu8.04.cfg"
This is probably not the method intended for using cobbler, but I was
running out of ideas.
This seemed to work, but now I am having an interesting problem
loading the mirror that cobbler created. In the preseed file I am
setting the mirror information up to use the mirror I created with
"cobbler repo add" like so:
### Mirror settings
# If you select ftp, the mirror/use_filesystem{ } filesystem{ ext3 } \
# mountpoint{ / }
\country string does not need to be set.
#d-i mirror/protocol string ftp
d-i mirror/country string enter United States
d-i mirror/http/hostname string 192.168.1.254
d-i mirror/http/directory string /cobbler/repo_mirror/Ubuntu_8.10
d-i mirror/http/proxy string
I know for a fact that the server is visible, as this preseed is being
loaded from that same IP address. I am confident, but not sure, that
the preseed file is being loaded because I was getting an error when
it was not finding it, and once I had a misspelling in it that raised
an error, and now I don't get any error messages. It just keeps
trying to connect to an external mirror.
Well, that was a lot of information. Hopefully someone has a suggestion?
Thank you so much for any hints you can offer me!
Jayson
15 years, 4 months
Quick question about snippets
by Joshua Baldock
Hey Guys,
Im not sure if this is possible. But is there any way to define in your kickstart file, that all the snippets in a particular directory get run? Rather then having to either list them individually in the kickstart or in another snippet?
Cheers,
Josh
________________________________
Information contained in this communication (including any attachments) is confidential and may be privileged or subject to copyright. If you have received this communication in error you are not authorised to use the information in any way and Optiver requests that you notify the sender by return email, destroy all copies and delete the information from your system. Optiver does not represent, warrant or guarantee that this communication is free from computer viruses or other defects or that the integrity of this communication has been maintained. Any views expressed in this communication are those of the individual sender. Optiver does not accept liability for any loss or damage caused directly or indirectly by this communication or its use.
Please consider the environment before printing this email.
15 years, 4 months
[devel] flock and NFS
by Michael DeHaan
Cobbler uses flock and /var/lib/cobbler/lock for it's
single-writer-at-a-time control right now.
NFS will not do flock.
It's not LSB correct but I propose this be "/etc/cobbler" since the
chance of /etc being on NFS are pretty much zero.
(Thinking aloud, is there any chance the two or so users that have
reported corruption of system records were using /var on NFS?)
Is there a better path for this?
--Michael
15 years, 4 months