I have a server that has 4 1.5 TB disks. Every disk has identical partitions

  • /dev/sd[abcd]1 200 MB
  • /dev/sd[abcd]2 2 GB
  • /dev/sd[abcd]3 1.5 TB (the rest of the disk)

/dev/sd[abcd]1 are RAID 1 device with four disks and it's mounted to the /boot partition. md3

/dev/sd[ab]2 /dev/sd[cd]2 are two different RAID 1 devices and both are swap partitions. md1 md2

/dev/sd[abcd]3 are RAID 10 device and it's the root(/) partition. md0

There is no installed LVM on top of them.

Now, I need to create a new 150 MB partition from md0 device. I have no direct access to the server but I may find a KVM, but it is not preferable as I have to use a terrible KVM device :) The server is newly formatted so the last option is to reinstall the system with the intended partitions and LVM on top of them to reduce future problems, but as I said I don't prefer to use it, since it will take 6-7 hours to install a new server. (and I have two identical servers)

What are my options?

  1. Reinstall the server
  2. Boot from CD and install LVM.
  3. ?

Thanks in advance,


"There is no installed LVM on top of them.". -- This is bad.

"it will take 6-7 hours to install a new server." -- I hope that includes racking, cabling, lunch, OS install, afternoon tea, and application configuration.

"I have two identical servers" -- then that makes it even easier.

  1. Reinstall, this time with LVM.
  2. Automate your OS installs (I'm going to guess you'll be needing kickstart, but if you're using a Debian-like system, you use debian-installer preseeding instead).
  3. Automate the installation and configuration of your application software, so that a full rebuild takes more like 30 minutes (elapsed -- about two minutes of that is manual time).
  • If you have to use a terrible KVM which is 12.000 km away from you, it takes that much time. Only writing an IP address takes up to 1 min, because the KVM stucks when you write or click something. Like if you write root, it becomes rrrrrrroooooooooootttttt, and if you delete a character it deletes all of them etc. BTW, it is Altusen KVM, beware of it! You have been warned. – Yuri Nov 29 '09 at 13:49
  • Automation means never having to use a shoddy IPKVM. – womble Nov 29 '09 at 20:31

Yeah, what womble said.


  • Why the reluctance to buy a proper hardware RAID card?
  • Why not combine /dev/sd[ab]2 and /dev/sd[cd]2 into a single RAID 10 instead of two RAID 1?

Better yet, why not create:

  • /dev/sd[abcd]1 as 128MB RAID 1 (md0) for /boot
  • /dev/sd[abcd]2 as the rest of the disk RAID10 (md1) as a LVM Physical Volume

Then create a volume group and create all the rest of the volumes from that VG.

  • These servers are powerful HP servers and have Raid Cards, but they are fake RAID cards (as the most of the RAID cards in the market), and in my experience Software Raid is more powerful than fake RAID cards, that's why I choose software RAID instead of those RAID cards. For the second question, Linux itself manages stripping between swap devices, there is no need to make it RAID 10, and the kernel is very good at this job. For the BOOT device, that was a mistake actually, but I did not mind to fix it, because it would cost more money. (100 mb disk area/30 min human power) – Yuri Nov 29 '09 at 13:47
  • If you've got HP servers, why aren't you using iLO instead of dodgy IPKVMs? – womble Nov 29 '09 at 20:31
  • Yes, thats a good question. I would want to use it more than anything. :) I love ILO, it is very fast and powerful. However, you have to subscribe and pay for it (and the evaluation licenses was expired long time ago). I made the suggestion and told them that it would save a lot of time, but there was a bureaucracy inside company, so getting the license would take more time and other people were waiting the server to be installed. bureaucracy == torturing myself :) – Yuri Nov 30 '09 at 3:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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