DRBL and Clonezilla

[Airbozo]

OK, quick note since I am swamped.

I am setting up a clone server based on Fedora 10, DRBL and Clonezilla to clone hard drives for a project. I will post up links later, but this is a place ot share knowledge and experiences.

I have a project to build 6 blade systems. Each Blade Chassis will hold 10 blades. Each blade is 2 complete nodes. Each node has 2 5670s and 2x 2gb (one for each processor) plus one 2.5" hdd.

All of these systems (120 in total) will be running Windows Server 2008 R2.

So I didn't want to install 120 copies of windows (we are not setup for network installs yet), and install and configure the customers software on 120 systems. Plus this is a rush order and I didn't want to camp out here for the next few days either. My solution? Install once and clone the drives. Since this is a volume license setup, All of the systems will have the same key. In fact everything will be identical until the SE sets it up at the customers site, then the hostname and IP address will change.

I have used the clonezilla live cd in the past and although it is slow, I found that it can be automated. Then I ran across the Clonezilla server tools. After a little reading I was able to setup this clonezilla server with minimal effort. In fact the hardest part was getting the base OS since our download speed at work sucks rocks.


I will post more info later since I am under the gun and must work on multiple projects at once, like the quad node servers (7 of them ) that go with this order running RHEL5v4 and a custom designed faceplate for another customer.. (more on that later).

[Oneslowz28]

While I will never have a need for cloning more than 2 or 3 drives, I am very interested in this. Can't wait for the additional post!

[Airbozo]

Update:

Still slammed, but I just tested a clone with 4 systems and the only extra time involved was waiting for the other clients to boot from the network. Once all systems are booted, then the image is sent out via multicast (other options are available, but I have not tested them yet). My only concern IS multicast since there is no verification. I am looking to see if I can run a verify command after the clone is done. One of the other options is unicast and then I think broadcast. I will be testing all of them to find the best solution.

This is great. My nightmares over the last week have been centered on how I can install verify and test 148 systems by friday... This takes care of the drive cloning, now I have to setup a multicast HD feed to test these boxes with known graded HD streams (4 of them ranging from grade 3 to 4.5).

[Luke122]

Back in the day, I used a Ghost multicast setup to deploy 120 systems at once; 4 classrooms. We disconnected the 4 rooms from the rest of the network, and ran all 4 to a dedicated server on a cart. We used PXE boot to pull the client setup, and had everything scripted. Once all the systems were connected, we pushed it out to all of them at once.

At the end of the push, the systems would reboot and run a script where we would type in the system number and room number. From that, it would rename the computer, and assign the proper IP address to each.

120 systems = 30 minutes, then 1-2 minutes at each to type in a few numbers. With 4 of us, we imaged them all and configured them in less than an hour.

[OvRiDe]

You might check out http://www.fogproject.org/. I just finished deploying VM's running FOG across our enterprise in order to handle imaging needs. I'm not sure if this will help your situation but it has saved me days and days.

[Airbozo]

Luke: I am doing exactly what you did, but for free... ;)

OvRiDe; Looks like we are doing exactly the same thing with different setups.

This setup so far has saved me 3 days of work. When I image the 6 blade chassis (120 full systems) it will save me weeks of work.

Still busy as hell and to top it all off we are doing our ISO 9001:2008 pre-assessment audit today and I seem to be the one with most of the answers... The real Audit is on the 2nd...

[x88x]

I've only used the live-CD version of CloneZilla, but I love it. In my experience at least, it's actually quite fast...especially compared to Ghost.. -_^ 15 minutes for an 80GB image off a USB drive is pretty good in my book. I'm very interested to find out how well the server version works on a large scale though. I seem to remember there is a limit on the number of targets it can do....but that may have been the unicast configuration...

EDIT:
Oh, and I think it goes without saying that if you can, pictures of this setup are required. :twisted:

[Airbozo]

I've only used the live-CD version of CloneZilla, but I love it. In my experience at least, it's actually quite fast...especially compared to Ghost.. -_^ 15 minutes for an 80GB image off a USB drive is pretty good in my book. I'm very interested to find out how well the server version works on a large scale though. I seem to remember there is a limit on the number of targets it can do....but that may have been the unicast configuration...

EDIT:
Oh, and I think it goes without saying that if you can, pictures of this setup are required. :twisted:

As far as I can tell, the "theoretical" limit is 253 clients per NIC card. The "working" limit may be lower, but since it is using multicast the bandwidth should be the same for 253 clients as it is for 1. I will be doing a test later today to see the impact of cloning 60 systems at once.

I completely forgot about taking pictures of the parts. Looking at 240 xeon's sitting on a cart was quite impressive. I had to shut down one of our SGI servers just to have enough power for all 6 of these blade chassis.

[DrkSide]

This is the same setup that I used to use when I did classroom setup for a training facility. I would make an image for whatever version machine ( we had a few difference levels) and whatever software was needed. Each time the class rolled around I would network boot the pcs and then push the image out. out of probably ~200 images I never had one fail once it started.

There is also a windows program that you can install that will respond to the clonezilla server and you can reboot and start imaging from windows. I wasn't around long enough to deploy the windows stuff but it always did interest me.

[simon275]

Nice work! Tools like clonezilla are a life saver. That many blades running windows... WHY!!!!!

[Oneslowz28]

Any pics yet Terry?

[Airbozo]

Any pics yet Terry?

Just brought in my camera to take pics before shipping.

I had a few issues, the main one being that once the cloned systems rebooted, they came up in the recovery mode. I just canceled out of that and the systems rebooted fine.

I had a couple of other issues, one with IP addresses. I had to add a static IP to all 120 systems... I also had to point all of the systems to the manager node as their time server. This is where I am lacking the proper windows admin skills. I know someone familiar with the powershell could have written a quick script to automate both tasks.

More info after I suss some coffee...

[Oneslowz28]

Good deal! I got the coffee flowing here this morning too. Already 3 cups in.

[Airbozo]

Forgot to post a few shots of the blades in action.

All 6 chassis running multicast tests;

http://img535.imageshack.us/img535/6162/dscf00430.jpg

A single chassis showing 10 blades with 2 nodes each;

http://img220.imageshack.us/img220/8639/dscf0046m.jpg

Each chassis has 10 blades and 2 1/10gb switch. Each blade is 2 complete nodes. Each node has 2 processors (E5640's) and 4x 2 gb Dimms, and one 300gb 2.5" hdd for a total of 120 processors and 240 2gb Dimms. Each node operates independently (not in cluster setup) running Windows server 2008 R2. What you cant see in these pictures is the manager node which collects data from each node and monitors the status of each node (health status). The manager node is a 1U system with 2 processors and 4x 2gb Dimms (basically the same specs as one node of each blade). There is also a windows XP test box running VLC streaming an HD video loop via multicast.

Notice all the power cords? I had to shut down several of our SGI and SUN supercomputers to power up and run the tests. In the short rack to the left in the first picture is my Clonezilla server (running Fedora 10) and 2 other test systems (one running RHEL 5.4 and one running Server 2008 R2).

[x88x]

Mmmm, yummy. :D That is an incredibly dense amalgamation of computing power. :D

[Airbozo]

Mmmm, yummy. :D That is an incredibly dense amalgamation of computing power. :D

I wish I had it here longer to play with. It is now at the customers site being setup for a demo...
\

[billygoat333]

mmm that is sexxxy!! lol. servers seem so complicated to me. *runs to read more*

speaking of clonezilla... is it a pain to ghost a drive to another using that program? I created a backup with it once, but haven't used it to copy to another hdd. seems like its fairly simple, I just haven't done it.

[x88x]

To copy an image to a drive, the process is just as easy as making the image in the first place. Just read the instructions and you should be fine.

[Airbozo]

mmm that is sexxxy!! lol. servers seem so complicated to me. *runs to read more*

speaking of clonezilla... is it a pain to ghost a drive to another using that program? I created a backup with it once, but haven't used it to copy to another hdd. seems like its fairly simple, I just haven't done it.


To copy an image to a drive, the process is just as easy as making the image in the first place. Just read the instructions and you should be fine.

Correct.

Clonezilla is really easy to use for creating images and restoring. It even works with Windows 7.

[billygoat333]

thanks all... I will report back tomorrow when my 2tb drive gets here so I can migrate my OS onto my 750gb drive :D

[Airbozo]

thanks all... I will report back tomorrow when my 2tb drive gets here so I can migrate my OS onto my 750gb drive :D

If your partitions are different size, make sure to read up on the specific section that talks about dealing with partition size mismatch.