-   Linux - Server (
-   -   RAID and backup layout re-jig (

DaveQB 05-17-2011 01:17 AM

RAID and backup layout re-jig
Hi all,

I have been running with a 3x750GB RAID 5 with 2 LV's on top of that. Been working well.
But I have now freed up 2 more 750GB's from another system and want to redesign this layout.

The aim is to have atleast 700GB as a main storage volume and shared out of NFS. Speed is important here (found the current RAID 5 slow on write speeds 19MB/s). And another volume to do incremental backups, so needs to be about 1.5-2x larger than the storage volume.

I also plan to have an offline backup box that powers up once a week to replicate the backup volume and then shutdown.

I have thought about this for a few days and have many different layouts written down, but I am not leaning to this:

2x750 RAID 1 = 750GB usable = main storage volume.
2x750 in a VG = 1.5GB = incremental backups

Then the incremental backups volume is backed up weekly to offline storage.
This leaves me with a spare 750 to start putting together this offline storage box.

I was thinking of having a 3x750 RAID 5 for backup, but figured what a waste of disk if I am going to have 3 copies of everything anyway, 2 of the backup volume.

I have thought of a big 5 disk RAID6, or a single disk storage and a 4 disk RAID6 for backup etc.

Can anyone see any problems with my final design?

zordrak 05-17-2011 03:55 AM

Bottom line for me is having offline backups.

In my home I have two identical servers. One is the primary server and it's configured for speed. The OS is on a RAID1, but the 6TB of storage volumes are not RAIDed. The second server only has one purpose:

1. Turn on.
2. Run rsync backups from server 1 with the -n flag.
3. Check the output of the dry runs looks good (nothing going missing etc).
4. Run rsync backups.
5. Turn off.

I can do this process whenever I want to backup the first server. The 1st server has basic redundancy so it stays up in case an OS disk fails - but if anything more than that fails then I have a backup server ready to go, or in the least a large number of spare disks, all ready to go with the data.

The point I'm making is that if you truly care about your data, redundancy (i.e. RAID) is of very little importance. The purpose of (non-0) RAID is not to back up data but to make it keep running if a disk fails. Without an offline backup you still have single points of failure in your RAID adapter, your filesystem, your LVM master data etc etc.

I'm not saying you don't realise this - you post clearly shows you realise the importance of backups - but once you have a backup configuration sorted out, you're just wasting resources putting RAID into the live configuration unless 100% uptime is really that important to you.

When I put together a proper storage system at work, then I did both. I had a RAID61 layout *and* a 4-depth tape backup system (nightly, weekly, monthly, quarterly). The reason for this was that the integrity of the data was massively important, but so was the availability. If the data was unavailable the company ground to a halt.

I'm getting off-point a little here I didn't mean to say so much.

I recommend that you first take a step back and decide: 1. What is your disaster-recovery strategy if every live disk goes bang. 2. How much money/resources do you want to put into making sure the server is always available no matter what. I would suggest that 1. is very important; 2 is probably not all that important so long as the data is safe.

DaveQB 05-17-2011 04:00 AM

Cool thanks.

About the offline backup...that's what I meant by:


I also plan to have an offline backup box that powers up once a week to replicate the backup volume and then shutdown.
And the not bothering with RAID for a live system is the conclusion I have come to but wanted to check with the LQ folks, thus my post.

Thanks for the reply.

zordrak 05-17-2011 04:14 AM

np :)

All times are GMT -5. The time now is 01:11 PM.