Tag Info

Hot answers tagged

60

To be honest, I think this comes down to simple viewpoint: Chef seems more of an imperative, programmatic solution, the usage of ruby as the language instantly makes me hope somebody ported it to python, as is the way of the world with all of ruby's ideas. That's not what you want for this sort of thing though. You want to speak to the void where the ...


19

I've written a detailed comparison of Chef vs Puppet here: Puppet vs Chef: 10 reasons why Puppet wins. Although it doesn't include use cases, I hope it provides some useful starting points for people wondering which tool to choose for their infrastructure automation.


14

Sorry about the verbosity. Use the tool that makes it's easy to get your job done. That's the point of automation, right? History: I have used puppet in past gigs and last month I spent about a week trying to get used to chef to see if I would make the switch at my new gig. I didn't leap. Jargon: One unfortunate problem with both of these systems is the ...


7

Yes, what you're suggesting is possible, and precisely what you'd use these tools in combination for. You'd use Clonezilla or Cobbler to push out the initial installs, and Puppet to keep the configurations in sync once they're installed. (And FYI, the other popular configuration manager for Linux you didn't mention is called Chef.) Clonezilla doesn't have ...


6

I disagree wholeheartedly with Kyle. If it is not necessary, it should be removed. It's a best practice to not install unnecessary software. The person undertaking the task, such as yourself, should be confident in the implications of the decisions they are making. Removing standard system utilities and libraries is generally frowned upon but that will ...


6

Have a look at the collection by David Schmitt http://reductivelabs.com/trac/puppet/wiki/CompleteConfiguration . When I don't know how to do something, I turn first to David's repo and see how he achieved it. A few modules I have used directly, but mostly I use it for reference. In practice with puppet you generally adapt other people's ideas and build your ...


5

The PXE ROM needs a "next-server" directive from the DHCP server in order to find and load the bootloader (be it grub, pxelinux or any other bootloader). If a "next-server" isn't supplied, it is up to the PXE ROM to decice what to do. You'll have to look at your network card bios configuration and see if there is an option to maybe specify the server ...


5

Here is a opinion: We have tried all of them in our company and we prefere puppet. Simply because it is easy to use.


4

The article in the Fedora wiki is probably one of the largest sources of information on kickstart options.


3

Here is a good blog entry on how to bootstrap puppet using cobbler. The author uses VirtualBox Virtual Machines, but you can easily adapt it to a physical server. http://number9.hellooperator.net/articles/2009/04/13/bootstrapping-puppet-from-cobbler Basically, you have a kickstarter server that runs puppetmasterd and cobbler. You customize a kickstart to ...


3

I don't believe there is a "standard base" of puppet. Based on my own experience, mailing list conversations, and analyzing many of the recipe collections on the internets, it appears everyone has different enough requirements that they just use the existing collections as examples. That's how my deployment has turned out.


3

What we do, is mount an ISO file we made, boot a kernel and initrd from that and have it load a kickstart from a central location. This kickstart file then point to a repository with RPM files, which could be your cobbler server. I haven't got much experience with Cobbler (sadly), but maybe this is an option for you?


3

The cobbler system is designed to manage the inventory and operating systems of many machines over time. While its feasible for you to manage your 1 pxelinux.cfg, think of a service provider environment where we frequently have to install different operating systems for customers. It can really improve the quality of life for the ops people who would ...


2

I think disabling things is a far better solution as a general rule. You start yanking stuff out and they might be dependencies of other packages, or other things might expect it to be there etc. All they will do after you disable them is take up a small amount of space and maybe add a little time to the updates. I think the idea is if you can you want to ...


2

It is being done right as it is. During the installation phase of a Xen virtual machine, the VM is booted in a slightly This is because of the way pygrub (not the actual grub) handles the booting of a virtual machine. We need to use pygrub because of the para-virtualized nature of Xen. Pygrub is used to 'feed' a kernel to the virtual machine. The VM boots ...


2

I'm doing this on our network. If your switch/router supports it, just relay the dhcp to the cobbler server with dhcp-agent or equivilent on the router. That way you don't have to have mulitple nic's etc on the cobbler server and make your life all confusing. Cobbler's default setup should do the rest once you have things up and running. Although you may ...


2

Generate the partitioning in the %pre section of your kickstart (which is simply a shell script). Dump it into a file, and then %include the file at the appropriate point in your kickstart. For example: %include /tmp/disk.ks %pre cat > /tmp/disk.ks <<EOF part pv.01 --size=1 --grow volgroup vg_$(hostname) pv.01 EOF Read more about %include ...


2

I've just discovered that /var/lib/cobbler has a .git respository, but it stopped getting commits about a month ago. All the commit messages are "API update" - presumably from the web interface. /etc/cobbler/settings contains scm_track_enabled: 1 scm_track_mode: "git" I find that "cobbler sync" on the command line generates a commit message of "API ...


1

Yes... and No.... Kickstart as part of it's installation process will configure Grub automagically. In addition you can manipulate the grub configuration files during the post install process. However it sounds like you are looking to set up a system which may involve multiple OS's which can be more of a challenge. There are may tools out there which can ...


1

You can also consider (R)?ex, it is easy to learn, and helps in configuration management and software deployment. It needs SSH or HTTP on his nodes.


1

Apparently, yes. Take a look at this Trac ticket on how to "allow templating variables to be used in --kopts". With regards to koan, the comments in the same ticket are informative. Relevant part reproduced here for completeness: This is implemented with @@variablename@@ for anything that runs through the templating engine (kickstart + PXE stuff). ...


1

I did this with NFS, try this: Install system-config-nfs and system-config-netboot (from yum) Start system-config-nfs and share some directory, lets say /srv/ Copy Centos Installer DVD content into /srv/Centos or just mount it there Start system-config-netboot and fill the inputs, in protocol type use NFS for the IP the server's IP where nfs service is ...


1

Probably all you need is to add url --url=$tree to /var/lib/cobbler/kickstarts/default.ks (or your own kickstart file). It will point to mirrored OS repo, see output of cobbler report distros for Kickstart Metadata line.


1

As I mentioned in this answer, redirect the SELinux rules to a file which is called cobbler.te. You can compile it with: # checkmodule -M -m -o cobbler.mod cobbler.te then create module package and install by executing: # semodule_package -m cobbler.mod -o cobbler.pp # semodule -i cobbler.pp


1

You may want to check out pxechain. PXEChain allows you to chain load pxeboot environments. I've seen this presented at some infrastructure conferences but never used it in production. I don't know if you can set a timeout to go to the next one or not but maybe worth exploring. ...


1

That's correct. Here's a bit more documentation if you need more info http://www.centos.org/docs/5/html/Installation_Guide-en-US/s1-kickstart2-packageselection.html


1

In a literal sense, no. The packages listed in the %packages section are installed by Anaconda, the Fedora installation program. This uses yum internally, but via API calls rather than exec'ing the yum command-line tool. That means the semantics of the install may differ slightly from what you get with yum install (particularly, the plugins configured will ...


1

Sure, you can just download them (with rsync or via any other method) and run them. They'll need to have the normal permissions any script might need. Are you trying this and encountering a problem?


1

I'm running Cobbler with Puppet in a pure Red Hat / CentOS shop. It's working fine, although you have to work out scalability issues with Puppet if you have several hundred servers. We were reworking the Puppet-templates for Xen/virt-install quite a bit, but the support for physical server installation with the Cobbler templates that are included with ...


1

if you just want to run some tests you can have several dhcp servers on the same network. you can have cobbler creating a dhcpd config that will tell a server only to respond when a particular MAC is making an ARP request. and if you simply comment out/disable any ranges on that server (and I am talking about isc-dhcpd here) it will not be intrusive. You ...



Only top voted, non community-wiki answers of a minimum length are eligible