up vote 2 down vote favorite
1
share [g+] share [fb]

I am interested in this as I want to set up my own headless servers, and I'm sure learning from google would not be a bad idea for me :)

link|improve this question
69% accept rate
feedback

2 Answers

up vote 7 down vote accepted

Be cautious in trying to follow Google too closely. The problems that Google solves are not the same problems that most IT professions do not worry about on a regular basis. Several items that Google probably does could help a lot of medium-sized sites, or even smaller installations. Some of these are guesses on my part. All all of this is heavily Linux centric:

  • Setup an automated, diskless method for staging new servers. Setup PXE on your server network to allow boot from network. Then use your distribution's method for automating installs, such as Redhat's kickstart. Assuming the servers are shipped with a blank hard drive, they should attempt to PXE boot, so installation can begin once they have booted.

  • Look at automated ways to configure your servers, once you have them up. Look at puppet or cfengine. This isn't needed for "headless" servers, but allows admins to interact less with the server.

  • Use SSH for a majority of your individual-server administrative tasks.

  • Of course, there are instances where physical console access is needed. If purchasing actual server-class hardware, make sure to include a remote access card (Dell DRAC, HP iLO, or whatever your vendor decides to call it). These cards generally have seperate power and network connectivity, and let you do almost everything over the remote connection.

  • For non-server hardware, setup console access through the serial port. A nice tutorial is here http://www.vanemery.com/Linux/Serial/serial-console.html It sounds a bit old fashion, but it is far easier to connect to a server through a serial port, then to drag a monitor and keyboard to it. I've personally done this more for systems out in remote locations, where monitors/keyboards are difficult to setup due to space constrictions.

link|improve this answer
+1 for the advice and the point that Google solves problems that are not typical in IT. They design their own hardware and maintain their own Linux distribution. Neither of these things scale down well. However all the points here I think scale down very well and and are good best practices to follow. Running production servers without some sort of change management like puppet is asking for problems. As is not automating the base installation. – 3dinfluence Feb 6 '10 at 16:00
feedback
up vote 4 down vote

i read some articles on them. It's basically red hat highly modified. Their server boot from lan, get the os, and then start to serve. So they don't really "install", but they are supposed to have around 800K servers, i guess you won't be near that number ;) (all the bad i wish you !)

resources:

http://www.cs.cornell.edu/projects/ladis2009/talks/dean-keynote-ladis2009.pdf

http://news.cnet.com/8301-1001_3-10209580-92.html

Google datacenter visit

One of the best resource: how to make your own google server like

link|improve this answer
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.