building a home server

M

Modder man

[H]ard|Gawd
Joined
May 13, 2009
Messages
1,770
i have been thinking about setting up a homeserver. i will probably start small and add on to it as needed.

i currently have a pentium d mobo and ram i figured i could use for a start

I got stuck when i started thinking about the drives and raid. mainly the raid card, i dont really see myself ever needing more than about 10 drives and would probably do something like raid 5. just kinda looking for any ideas you guys may have.
 
Last edited:
Skip RAID 5 with modern 2TB drives (smaller drives cost too much to attach to servers/ power). How technical are you? Think you can handle WHS? FreeNAS? Linux? Nexenta?
 
well im going to do this in part to learn this stuff, but i have used WHS why would you skip raid 5
 
10 drive RAID 5 = 20TB raw capacity. Let's say a 10^15 uncorrectable bit error rate for each drive. 1840GB * 1024 * 1024 *1024 * 8 = ~1.58 * 10^13 bits

(1 - (1.58 * 10^13 / 10^15) ) = (98.42% chance of a clean read of a single disk) ^ 9 drives left after a failure = 86.6% chance that you lose no data even IF you only lose 1 disk on a full array.

I'm a bit tired but that should be about right. That's why 2TB drives and RAID 5 don't mix well with lots of drives in the array. Swap to RAID 6 and that number goes WAY up.
 
Personally, I would stay away from RAID for long term storage. I've had RAID arrays fail 3 times times on me. One instance was a 2 drive failure on a RAID 5 array; a second drive failed during the rebuild process. After that, I just gave up on RAID. Everyone's experience is not the same, but just be warned.

Based on the feedback on the forums and things I have read online, it seems better to spend the money on a good Hardware RAID card and drives designed to be used in RAID. You can sometimes get away with cheaper drives, but there is still risk. Granted, I originally started messing with RAID years back on SCSI hardware, but it was my first failure and replacing it at the time as a student with a part-time job was difficult. I use WHS now and while it is not perfect, I have not lost any data yet (**knocks on wood**).

So yeah, be careful with pairing those large disks with a cheap controller. Better to do this setup right the first time.
 
yeah i kinda knew i wanted a hardware card but that doesnt help me a whole lot, i still dont know whats good or not, and or know enough about expanders and that kind of stuff.

drivewise i was kinda thinking of going hitachi, seems like thats what alot of you do
 
I would personally just buy two 2TB drives and rsync them, or whatever is available on your OS of choice. RAID isn't for backup, it's to minimize downtime. If you can live with your server being down for an hour or two, then I see no reason to pay for expensive RAID solutions.
 
but then if i lose a drive i lose it......or are you saying im better off just using a software backup to the drives
 
If you go hardware RAID you also need TLER-capable disks, or bad sectors will cause disks to be dropped from the array, degrading the protection that RAID offers considerably.

Why not just use WHS with duplication? No protection against BER but you need a backup anyway. So make two boxes or at least a real backup solution somehow; you can do that while avoiding RAID.
 
sub.mesa said:
If you go hardware RAID you also need TLER-capable disks, or bad sectors will cause disks to be dropped from the array, degrading the protection that RAID offers considerably.
Click to expand...

Still a small sample size (<50 disks), but I have literally not had this happen in the last two years using cheap consumer drives. I understand the mechanics of a drive's heroic recovery efforts which is why I am surprised to have not seen it. Then again, when a drive fails, I am usually happy to be using RAID 6 (and RAID-Z2) + hot spares.
 
Not sure how big your disks are, but especially with high-capacity disks today (2TB) the BER issue has become much more important. Perhaps you've been lucky, but most threads about broken onboard/hardware RAID arrays are about consumer disks who do deep recovery and the Windows/Hardware RAID engines can't properly handle those. As a result they 'fail' the disk, and write metadata to the remaining disks to 'remember' this event even across reboots and power cycles.

I've experiences this myself multiple times, including on my Areca ARC-1220 PCI-express hardware RAID controller (a 'real' controller with Intel IOP 500MHz). For the same reason a hardware controller is not suitable to ZFS. You can use it, but timeouts will still detach the disk from the OS; giving ZFS and FreeBSD no chance to be tolerant towards the drive. So you would want a real HBA on Linux/BSD. On Windows, TLER-drives are a necessity to avoid broken/split arrays. If you say you have 50 drives and not yet encountered this, consider yourself very lucky! Wait until one of your disks has a 'current pending sector' in the SMART output, indicating an active bad sector which is unreadable for the moment (exceeds ECC capability). It will cause the disks to spend more than 10 seconds in recovery cycle, after which the Hardware RAID controller considers this disk failed and updates the array to a degraded configuration. If you were already degraded or running RAID0 you now have a 'failed' array.

This can happen quite easily on 2TB disks in RAID5. One disk fails, so you buy a new one and initiate a rebuild, but then, one of the remaining disks have a small BER error, unable to read a sector. There you effort strands to rebuild the pool. Your data is not yet gone of course, but you do have corruption and a headache of getting it to work again. Also, many people who see their array as 'failed' will not know how to properly recover from that situation. Some forfeit their ability to recover and destroy and recreate their array.

RAID is a great technology, but it's implementations have been weak so far. Instead of increasing reliability and data security, it can effectively lower it by introducing a new single point of failure. The quality of the RAID engine and drivers/OS/fs is therefore very important. The only solutions really 'advanced' and 'good' in this regard, are ZFS and Btrfs; where Btrfs still requires a volume manager for RAID capability.
 
sub.mesa said:
Not sure how big your disks are, but especially with high-capacity disks today (2TB) the BER issue has become much more important. Perhaps you've been lucky, but most threads about broken onboard/hardware RAID arrays are about consumer disks who do deep recovery and the Windows/Hardware RAID engines can't properly handle those. As a result they 'fail' the disk, and write metadata to the remaining disks to 'remember' this event even across reboots and power cycles.

I've experiences this myself multiple times, including on my Areca ARC-1220 PCI-express hardware RAID controller (a 'real' controller with Intel IOP 500MHz). For the same reason a hardware controller is not suitable to ZFS. You can use it, but timeouts will still detach the disk from the OS; giving ZFS and FreeBSD no chance to be tolerant towards the drive. So you would want a real HBA on Linux/BSD. On Windows, TLER-drives are a necessity to avoid broken/split arrays. If you say you have 50 drives and not yet encountered this, consider yourself very lucky! Wait until one of your disks has a 'current pending sector' in the SMART output, indicating an active bad sector which is unreadable for the moment (exceeds ECC capability). It will cause the disks to spend more than 10 seconds in recovery cycle, after which the Hardware RAID controller considers this disk failed and updates the array to a degraded configuration. If you were already degraded or running RAID0 you now have a 'failed' array.

This can happen quite easily on 2TB disks in RAID5. One disk fails, so you buy a new one and initiate a rebuild, but then, one of the remaining disks have a small BER error, unable to read a sector. There you effort strands to rebuild the pool. Your data is not yet gone of course, but you do have corruption and a headache of getting it to work again. Also, many people who see their array as 'failed' will not know how to properly recover from that situation. Some forfeit their ability to recover and destroy and recreate their array.

RAID is a great technology, but it's implementations have been weak so far. Instead of increasing reliability and data security, it can effectively lower it by introducing a new single point of failure. The quality of the RAID engine and drivers/OS/fs is therefore very important. The only solutions really 'advanced' and 'good' in this regard, are ZFS and Btrfs; where Btrfs still requires a volume manager for RAID capability.
Click to expand...

I have maybe 5x 1TB drives, 20x 1.5TB drives and 25x 2TB drives (no clue how many but I have a full (Hitachi 2TB) Norco RPC-4220 and some others. Have been using Areca 1680LP, Adaptec 5805, and Adaptec 31605's for HW RAID thus far although the Nexenta based box has been using LSI 1068e's in IT mode and I'm testing a SAS 2008 based solution.

I totally understand the "heroic recovery" issue. And I will NOT use RAID 5 with 1.5TB+ disks. It isn't worth saving $90 over. Always double parity plus at least one installed hots pare. I am not saying I don't have failed drives, but double parity plus hot spares means that there is no downtime.
 
so if downtime does not really matter to me, all i want is to make sure my stuff is safe RAID is probably not the way to go?
 
if you want your data to be safe back the data up. Tape drives or even hard drives work well. i Hot swap HDD for back up. the drives are in a hidden safe. seems to work well. every 3 years i go through all the backup HDD and back them all up and i keep the new drives in the safe and store the old drives. HDDs are cheap. AS HDDs get bigger you need less HDDs.

once your DATA i SAFE it really don't matter. play with RAID if it fails and it will (expect things to go wrong as you learn) just go to your backups get the data and try again. RAID 5 wont work RAID 6 wont be a good idea as HDDs get larger too. RAID 6 works well now though.

i have been thinking of moving to a bunch of mini RAID 0 if one drive fails you loose your data but your data backup that should not be a big issue. with RAID 0 you get increase in performance and all space.
if say you use (4HDDs RAID 0) x 4 = 16HDD of space available and 1HDD on each RAID 0 would have to fail to have to rebuild all your data. RAID 0 may make your HDDs not last as long (so i read).
 
You must log in or register to reply here.
Back
Top