we have close to 50 servers. I would like to have X11 libraries (and devel). Instead of installing them one by one on all servers, I would like to have them in a central NFS location and have my app pick them up like that. Is it possible to do this with yum/rpm ?
On 16/09/12 22:29, Rita wrote:
we have close to 50 servers. I would like to have X11 libraries (and devel). Instead of installing them one by one on all servers, I would like to have them in a central NFS location and have my app pick them up like that. Is it possible to do this with yum/rpm ?
I've done this before on Solaris and regretted it. Worked well to start with, but turned into a World Of Pain.
It might be technically possible, if the RPMs are relocatable, or you did something cute with soft-links, but it will take a fair bit of mucking around, and will probably cause you bigger and harder to solve problems later on. Think about upgrading :-(
Might I suggest and alternative question: "How do I easily install and maintain X11 libraries on 50+ servers?".
Puppet, Chef, or Bcfg2 could do it but it will probably take you a long time to get your head around those beasts, let alone setting up the infrastructure.
A much simpler solution is ansible: http://ansible.github.com/.
Very easy to learn. Very easy to use. You can read _all_ the docs in under an hour. You should be able to set it up in a few minutes.
The following will ensure the latest version of xorg (and any dependencies) is installed on all my servers (updating/installing 10 at a time).
#> ansible all -m yum name=xorg state=latest -f 10
It can do a lot more, but that should solve your immediate problem.
Hope this helps,
Kal
On 09/18/2012 05:12 AM, Kahlil Hodgson wrote:
Puppet, Chef, or Bcfg2 could do it but it will probably take you a long time to get your head around those beasts, let alone setting up the infrastructure.
I dont know about that - it took me about 20 min to get my head around the basic 'types' in puppet and another 20 min to get the infra setup to manage about 5 machines ( this was when I was starting out with puppet ~ 2006'ish ? )
Am 18.09.2012 um 10:18 schrieb Karanbir Singh:
On 09/18/2012 05:12 AM, Kahlil Hodgson wrote:
Puppet, Chef, or Bcfg2 could do it but it will probably take you a long time to get your head around those beasts, let alone setting up the infrastructure.
I dont know about that - it took me about 20 min to get my head around the basic 'types' in puppet and another 20 min to get the infra setup to manage about 5 machines ( this was when I was starting out with puppet ~ 2006'ish ? )
your preconditions (skills) helped you to reduce the needed effort (efficiency). that must be normalized first to be comparable :-)
-- LF
On 18/09/12 18:18, Karanbir Singh wrote:
On 09/18/2012 05:12 AM, Kahlil Hodgson wrote:
Puppet, Chef, or Bcfg2 could do it but it will probably take you a long time to get your head around those beasts, let alone setting up the infrastructure.
I dont know about that - it took me about 20 min to get my head around the basic 'types' in puppet and another 20 min to get the infra setup to manage about 5 machines ( this was when I was starting out with puppet ~ 2006'ish ? )
I'm impressed :-)
I had a look at Puppet early this year, planning to migrate to it from existing Bcfg2 set up. There is a lot of good documentation, but there seemed to be a lot of options and it was not clear to me what the best path forward was. Reading a copy of Pro Puppet cleared things up, but I was still estimating the migration to take at least a week or two, with some hard thinking about relationships and structures.
K
On 09/18/2012 03:56 AM, Kahlil Hodgson wrote:
I had a look at Puppet early this year, planning to migrate to it from existing Bcfg2 set up.
Out of curiosity, what features motivated the migration? I use bcfg2, and the idea of using a system that doesn't feature reporting, interactive mode, or a "diff" from the running system is... difficult. Rather like trying to imagine using git or mercurial if they didn't support a "diff" command. You could still use them to manage your files, but would you want to?
On Thu, Sep 20, 2012 at 10:47 AM, Gordon Messmer yinyang@eburg.com wrote:
On 09/18/2012 03:56 AM, Kahlil Hodgson wrote:
I had a look at Puppet early this year, planning to migrate to it from existing Bcfg2 set up.
Out of curiosity, what features motivated the migration?
And has anyone looked at salt (http://saltstack.org/) yet? I like the concept of clients always being connected over an encrypted zeromq link, but haven't tackled anything more than running commands over it yet (which isn't really the right way to use it).
On 09/20/2012 09:08 AM, Les Mikesell wrote:
And has anyone looked at salt (http://saltstack.org/) yet?
I've looked at its documentation and talked to a few users. As far as I can tell, it has fewer features than bcfg2 -- including no "diff" function.
On 09/20/2012 04:47 PM, Gordon Messmer wrote:
Out of curiosity, what features motivated the migration? I use bcfg2, and the idea of using a system that doesn't feature reporting, interactive mode, or a "diff" from the running system is... difficult.
you should look at puppet, since it does all of those things and a bunch more - add in mcollective and theforeman to the mix, and you have quite a nice provision, manage and maintain environ ( I'd say add zabbix, while keeping an eye on sensu and extramon, in and you have an almost complete stack there ). The only bit that I think were still missing in the open source world is a good inventory management layer that interfaces upstream and downstream.
- KB
On Thu, Sep 20, 2012 at 5:07 PM, Karanbir Singh mail-lists@karan.org wrote:
you should look at puppet, since it does all of those things and a bunch more - add in mcollective and theforeman to the mix, and you have quite a nice provision, manage and maintain environ ( I'd say add zabbix, while keeping an eye on sensu and extramon, in and you have an almost complete stack there ). The only bit that I think were still missing in the open source world is a good inventory management layer that interfaces upstream and downstream.
I keep hoping someone will glue racktables and ocsinventory-ng together... Someone is at least working on a rest api for racktables. And you might like opennms for monitoring since they've added a provisioning concept to interface with an inventory database.
On 09/21/2012 12:02 AM, Gordon Messmer wrote:
On 09/20/2012 03:07 PM, Karanbir Singh wrote:
you should look at puppet, since it does all of those things and a bunch more
So I'd log in to a client managed by puppet and type what to see a 'diff' style report indicating how the puppet master would modify the system?
puppetd --test --noop
you can even partially apply manifests etc,
Also, we are likely offtopic here, the puppet lists are a good resource for these things
Am 18.09.2012 um 06:12 schrieb Kahlil Hodgson:
On 16/09/12 22:29, Rita wrote:
we have close to 50 servers. I would like to have X11 libraries (and devel). Instead of installing them one by one on all servers, I would like to have them in a central NFS location and have my app pick them up like that. Is it possible to do this with yum/rpm ?
I've done this before on Solaris and regretted it. Worked well to start with, but turned into a World Of Pain.
It might be technically possible, if the RPMs are relocatable, or you did something cute with soft-links, but it will take a fair bit of mucking around, and will probably cause you bigger and harder to solve problems later on. Think about upgrading :-(
Might I suggest and alternative question: "How do I easily install and maintain X11 libraries on 50+ servers?".
Puppet, Chef, or Bcfg2 could do it but it will probably take you a long time to get your head around those beasts, let alone setting up the infrastructure.
A much simpler solution is ansible: http://ansible.github.com/.
that looks interesting - thanks to share it.
-- LF
-
Gordon Messmer -
Kahlil Hodgson -
Karanbir Singh -
Leon Fauster -
Les Mikesell -
Rita