RAID 1 - Mirroring three drives?

[croc]

Not practical for a SOHO or SMB.... We do boot off of arrays at work, but they have dual fibre channel cards in the servers, and dual dual-channel controllers in the arrays... Expensive, but fast and reliable.

Most SOHO / SMB NAS solutions have a single channel ethernet port, thus not practical for speed / reliability reasons.

 

[belvdr]

Um, I meant that as a more humorous question, not reality... 😉

I never boot off arrays. I always use local disks for OS and the arrays for app and data storage. It's a habit I guess, as I see no additional benefit in converting.

 

[darkangelism]

tricky no, slow yes

 

[croc]

28 sec. start to finish, loading hpux, oracle 9i, initialising all the tables and taking new data is not really all that slow...

 

[darkangelism]

you running gigabit? on mine im thinking about using freenas and using dual nics and aggregate links to go 2 gigabit.

 

[bumster]

I prefer mirroring the OS on two small drives (the OS doesn't change that much) and then either mirror or RAID 5 separate data drives. This way you can image with Ghost or similar utility to an image file on a bootable DVD or image to a spare off-site drive. Then just backup your data to a USB attached drive(s) as often as you like. This will allow for a single drive to fail and still be up and running while also allowing for your disaster recovery with the off-site drives.

Also, on the data backup to the USB drives you may want to do tandem backups for taking off-site and leaving one at the office for quick recovery of corrupt or user deleted files without having to go get the backup drive from the other location. This makes the users very happy, instant restores.

 

[croc]

At home that wouldn't be a bad way to go, just make sure that the nics are capable of 'teaming' so that if you lose one the other takes the primary role. I still wouldn't use that as a boot drive, personally, as there are still too many SPOF's for a system that I have to depend on for 24 x 7 support.

The systems that I was referring to are typically HP 'N' class servers, with something like the HP MSA 1000/1500 drive array attached. The MSA units boot rather nicely over fibre-channel.

 

[darkangelism]

yeah i wont be booting from the NAS its just to backup, i am in the process of building a video server the motherboard is a tyan dual socket 771 with intel nics so they can be teamed and my switch is a Dell powerconnect 2708 web managed switch that does pretty much everything except jumbo frames. Though its probably excessive for a home file server its more fun they using consumer level stuff.

 

[peteb]

are looking for is called a tape drive. Carrying HDDs around and such is all very well, but not really the answer for backp purposes.

In addition as mentioned, plugging drives in and out of RAID arrays is still inherently risky and will seriously impact performance when you are doing a RAID rebuild (note it won't just update changes to the disk, it has to recopy block by block every bit state on the source drive)

 

[merald]

Heres the whole cake. I think I got the setup you trying to do.

Make a partition on the 3rd disk with the same size as in the mdraid.

run command

# mdadm /dev/md0 -a /dev/sdc3
where md0 is the raiddevice and sdc3 is a satadisk partition 3.

If you run the command

# mdadm --detail /dev/md0

You see something like this

--------------------------------------------
/dev/md0:
Version : 00.90.03
Creation Time : Tue Dec 26 19:16:41 2006
Raid Level : raid1
Array Size : 20972736 (20.00 GiB 21.48 GB)
Device Size : 20972736 (20.00 GiB 21.48 GB)
Raid Devices : 2
Total Devices : 3
Preferred Minor : 0
Persistence : Superblock is persistent

Update Time : Sun Jul 29 20:53:04 2007
State : clean
Active Devices : 2
Working Devices : 3
Failed Devices : 0
Spare Devices : 1

UUID : 367f461d:63371a04:8655586c:a2ec96bb
Events : 0.224

Number Major Minor RaidDevice State
0 3 3 0 active sync /dev/hda3
2 3 67 - active sync /dev/hdb3
1 8 35 1 spare /dev/sdc3

------------------------------

You see, that in case of faulty hda3 or hdb3, sdc3 will take their place.

Now simulate a faulty drive (either hda3 or hdb3 in my case) with command

# mdadm --manage --set-faulty /dev/md0 /dev/hdb3

What happens is that hdb3 sets to faulty drive and sdc3 will now be synct up with hda3.

---------------------
Number Major Minor RaidDevice State
0 3 3 0 active sync /dev/hda3
1 8 35 1 active sync /dev/sdc3

2 3 67 - faulty spare /dev/hdb3
----------------------

check progress with
# cat /proc/mdstat

When it synct up, just put it back

Remove the device
# mdadm /dev/md0 -r /dev/hdb3
Then add it again
# mdadm /dev/md0 -a /dev/hdb3

Now you have hdb3 as spare, if you want it back as it was, set the one you want to remove to faulty or just remove it with mdadm -r.

piece of cake

 

[merald]

offcourse you make a script that do this from cron a certain time every day or why not make a webpage-button @ the intranet "Sync backup" just ensure that the usb-disk will get the same devicename everytime...

 

[Crashman]

The same company that makes this also makes a 3.5" version:
http://www.tomshardware.com/2007/07/26/storage_accessories_for_geeks_a nd_pros/page2.html

There are also competing products.

You should be able to take two drives that are on a RAID 1 rack, and add them to the system as a single drive. Add another drive the same size as a single drive to the system. Use the system's RAID controller to do a RAID 1 between the 2-drive rack and the third drive. Make the third drive removable via a standard single-drive removable bay.

 

[willsmithorg]

To the original question, yes, here's how. You can choose between 2 disks in RAID1 with a hot spare that will only sync if something goes wrong with the original pair. Or you can have 3 disks in RAID1 (or more). Anybody who says otherwise is wrong, at least on a recent Linux kernel.


Start perhaps with a normal 2-disk RAID 1:

# cat /proc/mdstat
md1 : active raid1 sdb3[1] sda3[0]
87889848 blocks super 1.2 [2/2] [UU]
bitmap: 1/1 pages [4KB], 65536KB chunk


Duplicate partition table to new disk (in my case, sdd)

# sfdisk -d /dev/sda > /tmp/partinfo
# sfdisk /dev/sdd < /tmp/partinfo

Add sdd as a hot spare:

# mdadm --add /dev/md1 /dev/sdd3

Or, remove it as a hot spare, GROW THE RAID 1 to 3 DISKS, and add the new physical disk.

# mdadm --remove /dev/md1 /dev/sdd3
# mdadm --grow /dev/md1 --raid-devices=3
# mdadm --add /dev/md1 /dev/sdd3

After the resync,...

# cat /proc/mdstat

md1 : active raid1 sdd3[2] sdb3[1] sda3[0]
87889848 blocks super 1.2 [3/3] [UUU]
bitmap: 1/1 pages [4KB], 65536KB chunk


# mdadm --detail /dev/md1
/dev/md1:
Version : 1.2
Creation Time : Mon Sep 5 23:53:00 2011
Raid Level : raid1
Array Size : 87889848 (83.82 GiB 90.00 GB)
Used Dev Size : 87889848 (83.82 GiB 90.00 GB)
Raid Devices : 3
Total Devices : 3
Persistence : Superblock is persistent

Intent Bitmap : Internal

Update Time : Tue Sep 13 23:31:29 2011
State : active
Active Devices : 3
Working Devices : 3
Failed Devices : 0
Spare Devices : 0

Name : quad:1 (local to host quad)
UUID : 49249722:ca58f4eb:82802f4c:4bdc4954
Events : 1323

Number Major Minor RaidDevice State
0 8 3 0 active sync /dev/sda3
1 8 19 1 active sync /dev/sdb3
2 8 51 2 active sync /dev/sdd3

 

[Mousemonkey]

This topic has been closed by Mousemonkey