Raid controller failover?

J

Jeroen1000

Limp Gawd
Joined
Sep 17, 2010
Messages
266
Hi,

My home network is partially redundant and I've got an UPS for my "Home theatre room" which doubles as a surge protector. So just saying, I'm very interested in doing things right.

But the file server I'm building has almost no redudancy, apart from hooking it up to an UPS.

I'm thus wondering how redundancy is set up with regards to:

- Raid controllers
- Sas expanders
- Connections between RAID-controllers/expanders and hard drives.

Tried googling it but I can't dig up much useful info.

Cheers,
Jeroen
 
Not going to get failover without dual-port SAS drives or SATA/SAS interposers. It would be cheaper to build a second identical system for backups than to get failover working.
 
Hey Blue Fox, as usual you are right on the money. After reading for the better half of 3 hours I got the message that SATA drives won't cut it. I could financially cope with a second expander and RAID-card but not with the price tag associated with SAS drives.

Also Areca does not seem to support controller failover, nor does LSI. Bit strange as I thought them both to be top of the line controllers.
 
It's usually something reserved for a SAN. Most of mine have supported it (I actually have some of the interposer boards too). I thought LSI cards supported it, but I've honestly not worked with HA outside of iSCSI.
 
As BlueFox said, it would be cheaper to arrange for live replication. I have done this at work with identical pools and crossover 10GbE, with the synchronization tool depending on the environment. (e.g. DFS under Windows, manual cron job scripts under Linux.)

What you can do (that I've always done anyhow) is go redundant on your HBAs and expanders and make sure you have 2x capacity. In other words, if you are using a HBA, make sure you leave enough ports open that you can lose a controller and still stay up with some cable reshuffling while your controller is being replaced/repaired.
 
You can get some level of redundancy in raid 10 by using two controllers and assigning disks in mirrored pairs to alternate controllers. Obviously you won't get full performance with one mirrored set down, but your array should still be up and running.
 
I think I'll double check what RAID level I'll be using before putting any data on it.

This might best be asked in another thread (this has probably been asked before though) but...
I was thinking on going with 2 RAID 6 arrays with 12 drives each. That will amount to 20 * 2 TiB.

Before someone points out RAID is not a back up:D: The thing is I can't immediately buy all the hardware required for live replication. I had to safe up for what I have already.
So, I'm looking into whether RAID 6 is the safest level that exists.
I know of nested RAID levels but I do not know whether they offer more safety or not (requires more reading on my part). Also, there are quite some data recovery tools that can recover data from RAID 5 or RAID 6 arrays. Don't know whether those tools can make sense of nested RAID levels.

One final point I'm looking into, is there a way to combine 2 arrays on a software level? I'd like one big disk of 40 TiB instead of 2 smaller one's of 20 TiB each.
Ofcourse this wish may not increase chances of things going bad.
 
Jeroen, what are you after here, uptime, or just the guarantee that if the RAID card dies, you can get things operational again quickly? If it is the later, then I would just suggest a spare RAID card and expander put away in case your primary ones die.
 
Uptime: not so much. I thought my array would fail if the RAID controller died. Apparently, I can always replace it afterwards with no harm done (except downtime which is not really a big issue).

Speed: 250 MB/second is all I need (externally via ethernet that it).

Redundancy: yes please!

So I'm doubting between 2 RAID 6 array's (2 * 12 drives) or 1 RAID 60 array (1 * 24 drives).

Why RAID 6:

  1. Recovery tools will recognize this type of array = extra hope in case of faillure
  2. Rebuild times will be faster if a drive failes = reduces risk that even more drives will fail during a rebuild

Why RAID 60:
  1. 1 big blob of space. This is really useful and exactly what I want
  2. Problem 1: recovery tools might not work
  3. Problem 2: rebuild times will be longer and that might increase the chances of a faillure

Also , thanks to treadstone, I might be able to use flexraid to combine 2 RAID 6 array's in a view. This way it would look like I have one big blob of space while this is not really so.

So in short: I want to do everything humanly and $ possible to prevent faillure.
 
You must log in or register to reply here.
Back
Top