At least with large, rackmount, enterprise arrays, the general consensus with large drives is to use 2-drive parity, eg. RAID6. Depending on the load you put on the controller, there is a non-trivial chance that you'll have a second drive failure in the RAID-group during the time that you're rebuilding. Frequently a rebuild on either RAID5 or RAID6 with that big of a drive, might take 24-48 hours or more, sometimes longer. Two failed drives in a RAID6 doesn't lose any data, but in a RAID5 it would.
If you'll have a fairly light I/O load on the server, you might be able to get away with RAID5, since the reduced load will let the RAID controller (or kernel, for software-raid) spend more of its resources rebuilding when a drive does fail. Lloyd Brown Systems Administrator Fulton Supercomputing Lab Brigham Young University http://marylou.byu.edu On 06/22/2012 03:09 PM, Merrill Oveson wrote: > plug: > > Building a server: > > one 1 TB - for the OS and programs. > > four 3 TB raid 5 - which should give me 9 TB usable. Correct? > > How reliable are the 3 TB drives these days? Am I taking a big risk here? > > thoughts? > > > Thanks > > Merrill > > /* > PLUG: http://plug.org, #utah on irc.freenode.net > Unsubscribe: http://plug.org/mailman/options/plug > Don't fear the penguin. > */ /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
