Originating Author: Aaron Bowers

You've been asked to install, configure, and deploy a number of identical servers as quickly and efficiently as possible. What are your options? There are three main methods:

1. Build by hand
2. Script part or all of the install and configuration
3. Image the hard drives

This tip is a new twist on hard drive imaging and explains how to use hardware to clone hardware, instead of using software to clone hardware. I call it "RAID rebuild cloning".

RAID rebuild cloning

One way to address the shortcomings of disk cloning software is to use hardware-controlled RAID – we'll call it RAID-rebuild cloning. By design, RAID is used to create large volumes from smaller disks, provide fault tolerance, and increase performance of a disk subsystem. The most popular RAID levels are RAID1 or mirroring and RAID5 or disk striping with parity. If one disk fails in either RAID level the array will continue to function but in a degraded state until the faulty disk is replaced.

This same technology can be used to clone systems whose disks are very large (say 300 or more G-bytes), whose file systems are not recognized by hard disk imaging tools such as the LINUX EXT3 or XFS file systems, and whose system disks are hosted on a hardware-controlled RAID1 or mirror.

How it works

Here’s how it works -- In the event that a hard disk fails in a hardware-controlled RAID1, the controller has the ability to restore the entire contents of that failed drive to a replacement disk from the remaining functional hard disk. This is called rebuilding a degraded array. The disk you rebuild to becomes an exact copy of the disk from which the image is restored. The beauty of this is that the controller can rebuild the degraded array while the system is online and fully functional. And if your RAID controller supports automatic rebuilds, then starting this process is as simple as pulling out the failed disk and replacing it with a new one. The controller will detect the replacement and commence with the rebuild without any additional user intervention.

Now think about this: you have an exact copy of the system software, one on each disk. You can install each one of those disks in a system with identical hardware and one blank disk and rebuild that array as if it were degraded using your seed disk as the source of the rebuild. On our machines this rebuild process took around 3.5 hours; compared to nearly thirty hours just to create a software image of that system using hard drive imaging techniques. At the end of one round of rebuilds outside the original machine, you now have cloned two machines. If you want to clone a second round of machines, you may clone up to four at once. The number of machines you could clone at once grows exponentially, and each round of cloning takes just a few hours to complete rather than tens of hours or more with other techniques.

Conclusion

Although there are a number of techniques used to deploy a model system configuration to systems with identical hardware, some are much more time consuming, labor intensive, and prone to failure than others. Using RAID 1 members on our model LINUX system as seed drives for all the other RAID 1s in the same group of systems, proved to be an easy, rapid, and reliable method for system deployment.

To comment on this Professional Alert, please follow these simple instructions.

• Login before editing or commenting on this page.
• Comment on this article. You also can click the "+" tab above.
• Please sign your comments by typing "~~~" at the end of your comment.

Community Comments