Arch = x86_64 CentOS-6.4
We have a cold server with 32Gb RAM and 8 x 3TB SATA drives mounted in hotswap cells. The intended purpose of this system is as an ERP application and DBMS host. The ERP application will likely eventually have web access but at the moment only dedicated client applications can connect to it.
I am researching how to best set this system up for use as a production host employing RAID. I have read the (minimal) documentation respecting RAID on the RedHat site and have found and read a few online guides. Naturally, in my ignorance I have a bunch of questions to ask and I probably have a bunch more that I should but do not know enough yet to ask.
From what I have read it appears that the system disk must use RAID 1 if it
uses RAID at all. Is this the case? If so, is there any benefit to be had by taking two of the 8 drives (6Tb) solely to hold the OS and boot partition? Should these two drives be pulled and replaced with two smaller ones or should we bother with RAID for the boot disk at all?
Given that one or two drive bays will be given over to the OS what should be the configuration of the remaining six? It appears from what I have read that RAID 5 is the only viable option. It also appears that the amount of storage available on a RAID5 array with N members is N-1/N. I also read that as the number of members increase both latency and the risk of data loss increases. As the amount of disk space we have in this unit (24Tb) is greater than the total storage of all our existing hosts it appears that a RAID5 array of 5 units would leave at least one hot spare in the chassis and two if the OS is put on one disk.
Alternatively, the thought comes to mind that we could do a RAID1 with two RAID5 arrays each of which have 3 drives. Whether one would actually want to do that seems to me a bit questionable but it seems to be at least possible.
Comments, suggestions, caveats?
Regards,