Further questioning the RAID 1

[scott_417]

Not try to make it an issue :-) , just not so sure about this RAID 1. Assume for a server, Disk A's data is 100% duplicated on Disk B. In the event of the crashing of DIsk A, the Disk B will pick up the task. So the transition is smooth -- Peace of mind.

But, not for long, since I can not leave a crippled disk (Disk A) in the box, for what if the Disk B is someday crashed, then "who" will take over the task?. So I'll take the server offline to repair the Disk A, No?

If so, then what is the difference between using the tape-backupped data to do the recovery, and this RAID 1? Both eventually need to take the server offline for the repairing.

Therefore, using more disks to hold more data and do disk striping is more meaning to me, than using the RAID 1 to do the duplication.

How do you think of this, please share your view.



Thanks.

Scott

[Mike Bollman_1 · Accepted solution]

Scott,
There is no need to take the server offline to repair DiskA or the array. Simply remove the failed DiskA and replace with a known good drive. The array controller will rebuild the array on the fly without any downtime at all. Heck, if you have a hot spare configured it will even rebuild the array with zero intervention.

[Ivan Ferreira]

Restoring the data from a backup is slower that using mirror.

Even if you must take offline your server, you just replace the disk and boot. Then the data will be available and replicated to the replacement disk.

Stripping is only good for performance, but the availability of the stripping (MTBF) is the MTBF of the disk divided by the ammount of disks. So, if you have four disk in RAID 0, and the spected lifetime of the disk is 4 years, you will have a lifetime for the RAID 0 of 4/4=1 year.

If you loose 1 disk, all data is unavailable.

So, RAID 0 is good for performance but very bad for availability.