RAID and hard drive set, couple of questions.

[Steve Kalle]

Originally Posted by Scott Chichelli

hey if thats the way you feel by all means..
but its not what i recoomend or sell.. nor is your experiance the norm. we rarely have a client with data loss or raid 0 dying.
while raid 5,6 is the best way not everyone can afford it or wants to.
i have a hard enough time convincing them they need a back up plan at times.

FYI raid 1/10 i refuse to sell or support. take about thinking your butt is covered only to find out when you try to rebuild the array it fails... and talk about time? raid1 or 10 takes a good long time and you cant work whilst doing it..
you can copy over your back up vastly faster than trying to rebuild that array..
Scott
ADK

Your experience seems way off base! Raid 1 & 10 don't take much time at all to rebuild using the Intel ICH10r and it is as simple as copying data. There is NO parity data to calculate which is how R5 arrays go down and cause data loss. First, you say that you have no drives in R0 going bad but all of a sudden, drives in R1 and R10 do go bad?

My client with the bad sata port - his PC took all of 15 mins to rebuild the array and he could still use it while it was rebuilding; so, I really don't know where you get long rebuild times and an inaccessible computer. I used to use 4 Raptors in Raid 10 for my OS prior to Intel SSDs, and I lost 2 within a month, but I could still use the PC during each rebuild.

Talking about long rebuild times, then Raid 5 & 6 come to mind as they can take days.

[Scott Chichelli]

rebuild time vs copying your back up... i was refering to a comment you mad about how long it take to copy from back up...
:-)

Scott
ADK

[Randall Leong]

Originally Posted by Steve Kalle

Talking about long rebuild times, then Raid 5 & 6 come to mind as they can take days.

Yes, rebuilding a RAID 5 array can take days, especially on that onboard SATA controller. At least with a discrete PCI-e x8 RAID card, you can still work on the computer for other tasks that don't require the use of the RAID array while that RAID is rebuilding. The onboard SATA RAID controller, on the other hand, eats up a relatively high amount of CPU power during the rebuilding, which may prevent you from using that system at all during the rebuild.

[Randall Leong]

Originally Posted by Steve Kalle

Hi Randall,

I strongly disagree with this statement about compatibility with the Intel ICH raid. I have lost data due to the Intel raid dropping a WD drive from a Raid 0 array even though the drive was fine. The Intel software Raid cannot distinguish between a bad drive and a drive taking too long to respond. It drops the drive and a rebuild is required or data is lost if Raid 0. Also, a pro hardware raid controller can automatically reallocate bad sectors when writing to an array while the Intel software raid cannot. I became aware of this when my 3ware's alarm went off notifying me that a few bad sectors were being written to, so, it was able to hold the unwritten data in cache and find new sectors for this data. This happened about 3 years ago, so, I might not be remembering everything the 3ware software said. However, I recall investigating it and do remember that the Intel raid could not do the same thing.

I have investigated further, and found that using an enterprise hard drive with TLER (or CCTL or ERC) permanently enabled in a single-disk or a RAID 0 configuration can lead to data corruption or loss. All recently-shipping WD Caviar RE-series drives have their TLER permanently enabled and fixed at 7 seconds - and the TLER cannot be disabled at all whatsoever. That makes the drive suitable only for those RAID configurations which involves data redundancy, such as RAID 1, RAID 3, RAID 5, RAID 6 or a multi-level RAID.

I also learned the hard way that I had been getting corrupted data and all sorts of weird issues when I ran the Samsung F3 1TB hard drive with firmware version 1AJ10002. which has CCTL enabled and set to 7.5 seconds by default, as my OS drive. (Oddly, the 1AJ10002 firmware (March 2010) is actually older than the 1AJ10001 version (September 2010) that is im my other two Samsung F3 drives that I'm currently using in RAID 0; the 1AJ10001 firmware version supports CCTL but the feature is disabled by default.) Switching to a drive that either lacks TLER/CCTL/ERC or have such a feature disabled by default alleviated the problem. So, if I had to use a drive that has TLER/ERC/CCTL permanently enabled or enabled by default as an OS drive, I would have had to use two of those drives configured in a RAID 1 configuration as the boot/OS drive.

[Blane Nelson]

Great thread everyone!

So after my system drive crashed last Monday, I have rebuilt the system, and am ready to reconfigure the rest of the machine.

I am running one 500gb velocoraptor as the system drive, and 2 RAID 1s with 2 each HGST Deskstar 7K1000.C 0F10383 1TB 7200 RPM 32MB Cache SATA 3.0Gb/s hdds.

I am also running a 4tb (4x1tb drives) Glyph Forte RAID box via e sata that I use for backup and media storage.

I am looking for protection as much as I am for speed, so will this configuration work, or should I go back to R0 for the raid volumes?