Large desktop installation base management?

ponder

ponder

Honorary Master
Joined
Jan 22, 2005
Messages
98,657
Reaction score
31,319
Location
Niflheimr
Might have asked this before but can't remember. In a large corporate environment windows desktop are easy to manage via software tools that can deploy a HDD image via the network, automatically install software based on a user profile, keep updates goinge etc etc without end user support staff intervention or single desktop remote management.

How would one do this for say 20 000 pc's running linux? What tools are out there, commercial and OSS etc?
 
Last edited:
Red Hat provides pretty much everything you need in their Satellite server and related products. I don't work with this directly at work, and I work with servers rather than desktops, but at work we put a couple of hundred servers on-line every month, and so it's important that they're consistent. Kickstart gives us the ability to build a custom image, and with their software suite we can have a local RHN proxy and add our own custom channels to push pretty much anything we like out to all servers.

I don't know about OpenSource software to do the same, I work with RHEL almost exclusively at work.
 
Arthur said:
Acronis Snap Deploy deploys master images to PCs and servers. Not for updates - whole images only. With Universal Deploy, the same master image can be deployed to dissimilar hardware.
Click to expand...

Thanks but that's not what I'm after as it's only a single component of the whole thing which you can do with free software anyway.
 
I would think management software like this is very distro specific, e.g. Fedora/SL/RHEL/CentOS has things like Kickstart and Spacewalk, Ubuntu has Landscape, SUSE has Zenworks.

Now I concede that managing 20000 PCs is not the same as 20 to 30 in a PC lab, but I deploy images using Clonezilla with ssh. Yes it can do multicast. For package and config management I make use of rsync/pssh and some shell scripts that I wrote when I set up a small PC cluster at work. You could also use clusterssh or parallel instead of pssh. Bottom line is that whatever you use, you're going to have to customise it to suit your environment.

Edit:
As an aside, for monitoring I set up gkrellm to run as a daemon. This allows me to have a graphical, almost real-time view of what's happening on the cluster head-node, for example, which I VNC to. You could probably use munin or the more low-tech approach with mosshe as well.
For the lusers desktops and the PC lab I wrote a little script that runs as a cronjob and pipes some useful info into a webpage which is running on lighttpd on each machine. (This is also useful for accessing the cups admin page when the poor darlings can't print.)
Here's an example of the output I'm talking about:

Code: 
hostname                                                                                  
20h15 Friday 13/01/12                                                                           
Linux 3.1.5-pf x86_64                                                                           
2 x  Intel(R) Core(TM)2 Duo CPU     E6550  @ 2.33GHz                                            
[C1:43.0�C] [C2:36.0�C]                                                                         
93.54% of 1991MiB RAM used (128MiB free)                                                        
0% of 4000MiB swap used (4000MiB free)                                                          
Host up 7:10 3 users                                                                            
0.49 0.12 0.04 3/322                                                                            
/ 58% used (22G of 40G)                                                                         
/home 34% used (19G of 55G)                                                                     
[sda:T=38�C:PSect=0:BSect=0]                                                                    
[1] 5.3 root     bash            00:00:00                                                       
[2] 1.0 root     X               00:04:18                                                       
[3] 0.9 user     bash            00:00:00                                                       
[4] 0.7 user     conky           00:03:16

This is generated by a variation of a simple bash script which I also wrote for the cluster nodes because I then aggregate all this info into a web page on the head node so that I can see who's doing what.

Another edit:
Just found URLs for some more software related to the original query:
http://www.infrastructures.org/
http://www.linuxuser.co.uk/tutorials/puppet-server-management/
 
Last edited:
Thanks fskmh!

Puppet looks interesting. If you come across any other similar links please update this thread.

Ta.
 
Just read about a new release for this on Freecode (Freshmeat):

http://rexify.org/

What is (R)?ex

With (R)?ex you can manage all your boxes from a central point through the complete process of configuration management and software deployment.

Start exploring (R)?ex with this First Steps Guide.

We are always open for new suggestions and whishes. If you miss a feature just drop us a request.

Features

Integrates seamless in your running infrastructure.
No need of an agent. You only need a ssh connection to your server.
Simple definitions of tasks and profiles
Currently supports Debian, Ubuntu, Gentoo, Fedora, CentOS, Mageia, OpenSuSE and other Operating Systems. Read Compatibility List for more information.
Tasks are written in Perl, so everything is possible. See Perl Guide for a small introduction to Perl. (Just enough Perl for Rex)
Click to expand...

P.S. Clicking on the description tags for Rex yielded quite a few results in these categories:
http://freecode.com/tags/system-administration
http://freecode.com/tags/configuration-management
http://freecode.com/tags/deployment
 
Last edited:
 
For this kind of thing you wont only be needing tools to technically manage the computers because 20 000 computers will get out of hand. especially if multiple people supports it, they will end up doing their own things.

Standardization is the key.
You will also need software that involves HR, which can link Users and their Job descriptions to their computers and have access to only things they need. Integration changes and updates between I.T and HR should happen all the time.

Then in order to give proper support and have control over such a big network you will definitely need some sort of ticket based support system, With this you should also be able to see a mass problem before it happens by comparing support requests that come in.

As far as tools go to technically support Linux based systems. Like imaging HDDs with a standard OS image and remote support ect ect
Linux is a very flexible system and things can get very creative, lots of tools available, You can create your own "toolbox" packed with software to control 20 000 PCs from a central point. But i dont think you will get a Microsoft Domain with exchange, AD, DFS, SMS, SCCM WSUS ect ect that works nicely together with seamless password integration, which you can configure to go with the rest of your procedures.

a company who invests in 20 000 computers should spend money on a system that is proven to work.
and a Microsoft domain controller is by far the best (Dont get me wrong i like linux systems)
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X