New Server - RAID1 and Partition Recommendations Please!
 

Hey all! I've searched and read for days and I am almost more confused now than when I started. I am getting ready to build a new server (RHEL6) and I need some help with configuring the partitions and software RAID to best suit my needs.

I have three 2TB drives that I plan to use (qty 2) in a RAID1 config with the third disk being a spare. I also have a 250GB drive that I wasn't originally planning to use.

What I am trying to accomplish is to ideally have the entire server covered by the redundancy offered through RAID1. After reading though, I am confused as to where the /, /boot and swap partitions should be located.

Can I just create a single 2TB RAID1 partition at install and have everything on it and maintain true redundancy (one drive fails, the third will automatically take over)? Or if drive 1 fails, will the system fail to boot?

Would it be better to incorporate the fourth 250GB drive, make it drive 1 and just put /boot and swap on it and then put the /, /var, /usr, /home, etc. along with my other applications and data on the RAID1 partition? I realize this does nothing for system availability if drive 1 fails (though theoretically my data and core server data would be intact). Is it easier to replace drive 1 in that case and just create a new /boot and swap partition to get the server going? Does that even make sense?! See, I told you my brain was fried after all this reading.

Very confused here. Please help!!!

PP

If you are using SW RAID1 ie mdadm http://linux.die.net/man/8/mdadm, I'd go with 2 x 1TB as RAID1 (mirror) + 1 x 1TB Hot Spare and use the 250GB as extra non-critical data eg audio/video you can easily reload from eg Optical media or the net.
https://secure.wikimedia.org/wikipedia/en/wiki/RAID

Unless things have changed, you can only put /boot on a standalone disk or RAID1, not other RAID types or LVM.

NB: RAID only protects you from HW failure; if you delete something, its gone on both disks; you still need a proper Backup Soln.

Thanks Chris. I did read through the information contained in those links. What I am still unclear on is, if I just have a single 2TB RAID1 partition with all of my data on it (including /boot) and one of the drives fail, will the spare automatically be rebuilt with the data (assuming I have it configured to do so) and thus the system will boot properly without any further intervention on my part? Or, will I have to do something manually (after the drive failure) to make this happen (i.e. - is the MBR only written to /dev/sda at install)?

Also, and I welcome any and all feedback here, given the selection of drives I have available for this build, I would love to hear how others would go about configuring RAID and/or the partitions.

Thanks again!
PP

Hey Pix,

This is what i would do,
RAID1 with 2x2TB, partition:
/boot - 100MB, insure the grub is installed on /dev/sdb as well (in case /dev/sda will be the on to fail first)
/tmp - 1G
/swap - 2xRAM (if you got 4G of RAM then 8G swap)Keep in mind swap wont be on RAID but rather have 2 partitions on each drive (4G each in this example)
/ - rest

Wouldn't use the 250G as it will ad to CPUs overhead. As for the third one you can keep it as spare in case of of them fails, you'll have to manually rebuild it (no biggie just few commands) http://www.howtoforge.com/replacing_..._a_raid1_array

Not to mix you up even further, have you considered RAID5? Altough, from my experience software RAID5 isnt the best option.

Peace

Well, RAID5 is an idea, but min is 3 disks https://secure.wikimedia.org/wikipedia/en/wiki/RAID so if one dies you'll need to nip out quick and buy another one before the RAID fails.
RADI1 (mirror) plus hot spare gives you more safety (timewise) at the expense of less storage space; its a trade-off, your call.

You don't need 2 X RAM swap any more, that's an old calcn; I'd go with 1 x RAM, maybe 1.25. Really depends on how hard you want to push the system.
Of course, disk space is very cheap compared to RAM these days ...