Hi all, 1. Is RAID-5 fault tolerant? 2. I make a RAID-5 array with 3 disk. The /boot is on RAID-1 partition, the rest are raided 5. When I simulate disk failure by unplugging sda and restarting Centos. Grub fails in stage 1.5. What is the correct way of simulating harddisk failure in RAID-5? If I unplug the other drives, it manages to boot ok. Thank you.
Hi, once you remove sda, you need to setup grub on sdb to get it working. Just boot from the CentOS cd1 into linux rescue mode, and it should autodetect your system and mount it under /mnt/sysimage. Once you're in, chroot /mnt/sysimage /sbin/grub --> you get a prompt grub>
In my case I read in my grub.conf: title CentOS (2.6.9-22.0.2.106.unsupportedsmp) root (hd0,0)
so at the grub> prompt I would type
grub> root (hd0,0) grub> setup (hd0) grub> quit
You can then exit, reboot and your system should be up and running.
Hope this helps
Fajar Priyanto wrote:
Hi all,
- Is RAID-5 fault tolerant?
- I make a RAID-5 array with 3 disk. The /boot is on RAID-1 partition, the
rest are raided 5. When I simulate disk failure by unplugging sda and restarting Centos. Grub fails in stage 1.5. What is the correct way of simulating harddisk failure in RAID-5? If I unplug the other drives, it manages to boot ok. Thank you.
On Thursday 09 March 2006 05:07 pm, Simone wrote:
Hi, once you remove sda, you need to setup grub on sdb to get it working. Just boot from the CentOS cd1 into linux rescue mode, and it should autodetect your system and mount it under /mnt/sysimage. Once you're in, chroot /mnt/sysimage /sbin/grub --> you get a prompt grub>
In my case I read in my grub.conf: title CentOS (2.6.9-22.0.2.106.unsupportedsmp) root (hd0,0)
so at the grub> prompt I would type
grub> root (hd0,0) grub> setup (hd0) grub> quit
You can then exit, reboot and your system should be up and running.
Hope this helps
Thanks Simone, It works now. Anyway, can you give me an insight how RAID-5 works in preventing the loss of data?
I'm testing it by copying a large file (20MB) into /root, and when I unplugged a drive, it seems that the data is still Ok. How does RAID-5 do this?
Thank you,
Fajar Priyanto wrote:
On Thursday 09 March 2006 05:07 pm, Simone wrote:
Hi, once you remove sda, you need to setup grub on sdb to get it working. Just boot from the CentOS cd1 into linux rescue mode, and it should autodetect your system and mount it under /mnt/sysimage. Once you're in, chroot /mnt/sysimage /sbin/grub --> you get a prompt grub>
In my case I read in my grub.conf: title CentOS (2.6.9-22.0.2.106.unsupportedsmp) root (hd0,0)
so at the grub> prompt I would type
grub> root (hd0,0) grub> setup (hd0) grub> quit
You can then exit, reboot and your system should be up and running.
Hope this helps
Thanks Simone, It works now. Anyway, can you give me an insight how RAID-5 works in preventing the loss of data?
I'm testing it by copying a large file (20MB) into /root, and when I unplugged a drive, it seems that the data is still Ok. How does RAID-5 do this?
Thank you,
It uses parity information that it stores across all three drives. In your setup 1/3 of each drive is used for the parity information. Basically any two drives then have enough parity information to recreate the date on the third drive. With raid 5 the usable storage is C * (N -1) where C is the capacity of the drives and N is the number of them. So if you have 3 * 250GB disks the usable storage is 500GB. If two disks die at the same time kiss all your data good bye (although this is a pretty rare scenario)
On Thursday 09 March 2006 06:42 pm, Nick wrote:
It uses parity information that it stores across all three drives. In your setup 1/3 of each drive is used for the parity information. Basically any two drives then have enough parity information to recreate the date on the third drive. With raid 5 the usable storage is C * (N -1) where C is the capacity of the drives and N is the number of them. So if you have 3 * 250GB disks the usable storage is 500GB. If two disks die at the same time kiss all your data good bye (although this is a pretty rare scenario)
What I don't understand is that when it's running with only 2 disks out of 3, it seems that it's still running OK.
So: 1. What is the consequences when RAID-5 array runs with 1 lost disk? Will it damage the data? 2. What will happen if I copy a large data (say 30MB) into the running array with 1 lost disk?
Fajar Priyanto wrote:
On Thursday 09 March 2006 06:42 pm, Nick wrote:
It uses parity information that it stores across all three drives. In your setup 1/3 of each drive is used for the parity information. Basically any two drives then have enough parity information to recreate the date on the third drive. With raid 5 the usable storage is C * (N -1) where C is the capacity of the drives and N is the number of them. So if you have 3 * 250GB disks the usable storage is 500GB. If two disks die at the same time kiss all your data good bye (although this is a pretty rare scenario)
What I don't understand is that when it's running with only 2 disks out of 3, it seems that it's still running OK.
So:
- What is the consequences when RAID-5 array runs with 1 lost disk? Will it
damage the data? 2. What will happen if I copy a large data (say 30MB) into the running array with 1 lost disk?
Running with 1 disk down the array will function at 100% as all the data is still available because of the parity info stored on the two good disks. The problem is with one disk down you have absolutely no redundancy. If one of the other disks failed the data would be lost for ever. If you replace the dead disk you can then start rebuilding the array so that you get your redundancy back.
this page explains it a lot better than i can http://www1.us.dell.com/content/topics/global.aspx/power/en/ps2q03_luse?c=us...
On Thursday 09 March 2006 07:12 pm, Nick wrote:
Running with 1 disk down the array will function at 100% as all the data is still available because of the parity info stored on the two good disks. The problem is with one disk down you have absolutely no redundancy. If one of the other disks failed the data would be lost for ever. If you replace the dead disk you can then start rebuilding the array so that you get your redundancy back.
this page explains it a lot better than i can http://www1.us.dell.com/content/topics/global.aspx/power/en/ps2q03_luse?c=u s&l=en&s=corp
Thank you very much Nick. It really gives me a clear understanding.
Jeffrey Tadlock wrote:
Nick wrote:
If two disks die at the same time kiss all your data good bye (although this is a pretty rare scenario)
Just had this happen last week! First time in a long, long time though, so I certainly agree with it being a rare scenario.
Time to upgrade to raid6! ;)
Jeffrey Tadlock spake the following on 3/9/2006 5:41 AM:
Nick wrote:
If two disks die at the same time kiss all your data good bye (although this is a pretty rare scenario)
Just had this happen last week! First time in a long, long time though, so I certainly agree with it being a rare scenario.
/jt
Usually happens with some sort of electrical surge, fire, flood, stupid mistake, or other such catastrophy. Raid does not remove the need for backups, it just makes the system more reliable. If your server room bursts into flames, no amount of raid, short of some kind of mirror over iscsi or drbd to a server in another physical location would save you then.
Scott Silva wrote:
Raid does not remove the need for backups, it just makes the system more reliable. If your server room bursts into flames, no amount of raid, short of some kind of mirror over iscsi or drbd to a server in another physical location would save you then.
Agreed. We had full current backups and were able to restore with no issue.
/jt
-
Fajar Priyanto -
Jeffrey Tadlock -
Nick -
Scott Silva -
Simone