Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I'm building new servers for a project I'm working on. They will all run Ubuntu Server x64 (10.04 soon) and require a RAID 1 hotswap configuration (just two drives) to minimize downtime.

I'm not worried about Raid performance. The server hardware will have plenty of CPU power, and I'm only doing a RAID 1. My only requirements are:

  1. Everything, including the OS, must be mirrored.
  2. There must be no down-time when a drive fails. I need to be able to swap out the failed drive with another and have the RAID rebuild itself automatically (or maybe by running a simple script).

I'm wondering if the built-in Ubuntu Software RAID can handle this, particularly the hotswap part. 10.04 looks promising.

I'm considering buying the 3Ware 9650SE-2LP-SGL RAID controller, but with the number of servers we're purchasing, that would increase the total price quite a bit.

Any advice at all would be appreciated. Thank you.

share|improve this question

3 Answers 3

up vote 4 down vote accepted

I have hot swapped drives using the software RAID builtin into the Linux kernel on many occasions. You may need to run a command to add the new device. I believe it is possible to make it automatic, but in the places where I use it manually running the command to add the new drive has never been a problem.

I am not entirely certain that the computer will survive with zero downtime. That may depend on your hard drive controller and how it responds to a drive failure.

share|improve this answer
1  
+1 For actually answering his question :-) –  Kyle Brandt Mar 5 '10 at 20:43
    
Thanks for that answer. So what I gather from what you're saying is that to guarantee zero downtime, I would need the hardware RAID, but it could be possible with software RAID depending on the motherboard's drive controller. Is that about right? –  Andrew Mar 5 '10 at 20:48
2  
In one HD failures I have seen a HD go bad in a way that the hardware RAID controller got messed up to the point that the OS crashed and a reboot was required. But that happened several years ago. I haven't seen that happen recently. –  Zoredache Mar 5 '10 at 21:05
    
Ok, so maybe not guarantee, but close enough. –  Andrew Mar 5 '10 at 21:07
    
Also, do make sure you verify that the motherboard controller supports hot swapping. –  Zoredache Mar 5 '10 at 21:08

I think the other posts have answered the question but I have a somewhat related thought to add.

Since uptime is important in this application. Make sure you're using Puppet and Kickstart for setting up and maintaining the configurations on the servers. Also make sure you have a good backup solution....rsnapshot works pretty good.

The hardware should be pretty replaceable cogs once you're dealing with any sort of scale of computers. Because you'll eventually have to deal with the following situations....you need a plan on how you're going to deal with them now. Not when it happens.

  • Even with redundant power supplies, raid, etc machines will fail in time.
  • The situation where a client starts to outgrow the hardware they are on....if all the clients are on separate hardware as your dialog to some of the answers to seem to imply.
  • Hardware replacement. In 5 years or so you'll want to replace hardware.
share|improve this answer
    
Thank you for that advice! I had never heard of Puppet or Kickstart before. I've written my own Bash scripts for setup, but looking into those now. –  Andrew Mar 6 '10 at 19:25
    
Also look at preseeding which I believe is the official Debian/Ubuntu way to do it. Although I believe that kickstart also works. wiki.debian.org/DebianInstaller/Preseed –  3dinfluence Mar 6 '10 at 22:47

Are you sure you need RAID?

I only ask because you said with the number of server you are purchasing, that controller will increase the cost a lot. So with this many server for a single project, aren't a lot of these servers redundant? Perhaps you would save more money by not getting all those second hard drives?

share|improve this answer
    
Each server will need to be highly available. If at all possible, we need to avoid even one of them being down for an entire day while it's being rebuilt. –  Andrew Mar 5 '10 at 20:45
    
Ah, well then, you are sure :-) –  Kyle Brandt Mar 5 '10 at 20:47
    
You posed a valid question. I did say any advice would be appreciated. :-) –  Andrew Mar 5 '10 at 21:00

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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