Home NAS: RAID recommendations and question about IDE drives

Forums › Linux › Home NAS: RAID recommendations and question about IDE drives

foobar

186 posts

Master Geek

# 30239 2-Feb-2009 15:14

Hello!

I'm putting together a small NAS box for our home network. It would mostly be used to store photos, videos and backups from our laptops. It's an old PC we still had sitting around and which now hopefully can be put to some good use. I'm not much of an expert as far as drives are concerned, so I have a few questions.

Should I go with IDE drives? As I said, the box is a bit older. I read somewhere that you can only attach two devices per IDE controller, and it might just be that the box just has one of those. There is the drive from which it boots (xubuntu, by the way) and then I was thinking about adding two more drives in RAID 1 configuration. But if only two devices can be attached to the controller in total then this wouldn't work, right?

Here is the output of lspci:

00:00.0 Host bridge: nVidia Corporation nForce2 IGP2 (rev c1)
00:00.1 RAM memory: nVidia Corporation nForce2 Memory Controller 1 (rev c1)
00:00.2 RAM memory: nVidia Corporation nForce2 Memory Controller 4 (rev c1)
00:00.3 RAM memory: nVidia Corporation nForce2 Memory Controller 3 (rev c1)
00:00.4 RAM memory: nVidia Corporation nForce2 Memory Controller 2 (rev c1)
00:00.5 RAM memory: nVidia Corporation nForce2 Memory Controller 5 (rev c1)
00:01.0 ISA bridge: nVidia Corporation nForce2 ISA Bridge (rev a4)
00:01.1 SMBus: nVidia Corporation nForce2 SMBus (MCP) (rev a2)
00:02.0 USB Controller: nVidia Corporation nForce2 USB Controller (rev a4)
00:02.1 USB Controller: nVidia Corporation nForce2 USB Controller (rev a4)
00:02.2 USB Controller: nVidia Corporation nForce2 USB Controller (rev a4)
00:06.0 Multimedia audio controller: nVidia Corporation nForce2 AC97 Audio Controler (MCP) (rev a1)
00:08.0 PCI bridge: nVidia Corporation nForce2 External PCI Bridge (rev a3)
00:09.0 IDE interface: nVidia Corporation nForce2 IDE (rev a2)
00:1e.0 PCI bridge: nVidia Corporation nForce2 AGP (rev c1)
01:06.0 Ethernet controller: Intel Corporation 82557/8/9/0/1 Ethernet Pro 100 (rev 0c)
01:08.0 Modem: PCTel Inc HSP56 MicroModem (rev 04)
01:0b.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8169 Gigabit Ethernet (rev 10)
01:0c.0 RAID bus controller: Integrated Technology Express, Inc. IT/ITE8212 Dual channel ATA RAID controller (rev 10)
01:0d.0 RAID bus controller: Silicon Image, Inc. SiI 3112 [SATALink/SATARaid] Serial ATA Controller (rev 02)
01:0e.0 FireWire (IEEE 1394): Texas Instruments TSB43AB23 IEEE-1394a-2000 Controller (PHY/Link)
02:00.0 VGA compatible controller: nVidia Corporation NV18 [GeForce4 MX 440 AGP 8x] (rev a2)

Does that look like I have only one IDE controller? There is an IDE bridge and an IDE interface listed...

Could I use serial ATA drives? Or SATA even? I assume I need to get a special controller card for that?

Thank you very much...

tonyhughes

Hawkes Bay
8477 posts

Uber Geek

Mod Emeritus

Trusted

Lifetime subscriber

# 193428 2-Feb-2009 15:21

Being that I/O is not a factor in your situation, I would be inclined to recommend getting a cheapie 4 port SATA card, and either using onboard RAID 1, or software RAID 1 in Linux.

foobar

186 posts

Master Geek

# 193453 2-Feb-2009 16:42

tonyhughes: Being that I/O is not a factor in your situation, I would be inclined to recommend getting a cheapie 4 port SATA card, and either using onboard RAID 1, or software RAID 1 in Linux.

Thank you for the reply. Exactly right: Raw I/O performance is not much of an issue here. As far as the 'cheapi 4 port SATA card is concerned', are you talking about something like this here? That's a $49 SATA card from Dick Smith. It talks about WIndows support in the item description, so I'm wondering of course if that would work out of the box with GNU/Linux. It doesn't seem to mention the exact controller type.

Are there other options for buying Linux compatible SATA cards? I mean, Dick Smith can't be the only place, right? Which other places can you recommend?

Thank you again...

rphenix

894 posts

Ultimate Geek

Subscriber

# 197877 24-Feb-2009 23:08

foobar:
Are there other options for buying Linux compatible SATA cards? I mean, Dick Smith can't be the only place, right? Which other places can you recommend?

Thank you again...

Just use pricespy, find out what chipset the controller is and then google it to see how well supported it is.

Viddy

7 posts

Wannabe Geek

# 197890 25-Feb-2009 01:17

Heh, that dmesg output looks very familiar - its what I've made my fileserver out of - a nforce2 board with a amd xp something something on it.

The silicon image onboard sata chipset works under a recent version of the kernel - ubuntu 8.04+ - where works is hotpluging and booting/reading from.

I've had quite bad experiences with the alternate ide controller, with data corruption:
01:0c.0 RAID bus controller: Integrated Technology Express, Inc. IT/ITE8212 Dual channel ATA RAID controller (rev 10)

So, my configuration is two independant raid sets, running linux software raid for everything.
The old raid array is 4x80gig PATA (parallel ata) array hanging off the first two ide ports, and two off a Promise Technology, Inc. PDC20268 (Ultra100 TX2) pci card
The other array is 5x 320G sata drives hanging off a 4port Adaptec Serial ATA II RAID 1420SA, and 1x320gig off a onboard sata port.

This enclosure holds 4 out of the 5 sata drives http://www.pbtechnz.com/?item=IDEMRA500

And this http://www.pbtech.co.nz/index.php?item=CASCLM13046 holds the 4x pata drives. Fans are good, because keeps the drives from getting too hot.
A hotswap enclosure is also good, just so you don't have to take the case apart if (when) a drive fails, even if you don't actually hotswap the drives.

For your situation, I'd go for two new sata drives with reasonably different serial numbers and different date codes, running raid 1, with the minimum number of platters as they run cooler, and less moving parts can fail (see the manufacturers websites for platter counts). I'd also go for seagate drives, but thats a personal choice.

If you do go the parallel ata (PATA) route, do one drive per port - don't do a drive on master and slave, as ide only allows for one drive to talk at a time on a cable, so your effective bandwidth gets halved, but theres something else that makes performance in that configuration entirely crap. Don't remember what it is.

You're also going to run into the interesting problem of choosing if you're going to run your operating system on your raid, or some other method. I've made up a usb key, and booted off that, but its reasonably technical, and my board has some issues around warm booting off usb.
Booting off raid 5 is a problem, but not raid 1, although it needs to be carefully investigated, and tested. (you want both disks of the mirror to be the same, right?)
The reason that you may want to consider running your os on the array is that you've got an extra copy of the operating system ready to go in the event of a drive failure, and you won't get as much down time.

The drivers that get included with cheap (read here less than ~$500) raid cards, that say that they do onboard raid - they lie. They mark the drives in the card bios, then get you to load a binary blob in to the kernel that doesn't work as fast as the linux software raid drivers, then breaks horribly when you do a kernel upgrade, or distro update, and the drive contents aren't even vaugely compatable with anything else out there.
Linux software raid, however, is in my experience, bloody good. It also seems to be backwards compatable. You update the kernel, and it detects the array, and all is good, and everything works like it should.

Filesystems - stay away from reiserfs, though I don't know that its an option anymore - it seems to quietly corrupt data. I use and have used jfs for about 4 years now, and it seems to be good. No apparent data loss - the legal mp3s that have been there still play. We also use xfs and ext3 at work, depending on the system requirements.

Seriously consider a UPS, if only to avoid reboots where you have to manually fsck the array. You can get around this by using something like jfs, but even then it will eventually need some gentle fscking, given enough hard poweroffs.

Run smartmon, and mdadm in daemon mode to firstly check the smart status of the drives, and secondly to alert you to when the array gets into degraded mode because a drive has failed. Its not unheard of for a mirror array to have a drive fail, and everything keep on working, and no-one notice till the second drive fails :)

Thats about it for my brain dump. The devil is, of course, in the implmentation. Good luck.

barf

643 posts

Ultimate Geek

# 198931 3-Mar-2009 11:37

You may save some time by going for a distro like OpenFiler.

Personally, I'd stay away from RAID in favor of LVM unless this is server kit with Host Bus Adaptors.

LVM is the goods. It provides more flexibility than RAID and retains redundancy, definately worth a look since it also facilitates block-device level encryption, snapshots and abstracts the storage model to provide Volume Groups that can span disks of any size (no more wasted bytes to RAID limitations). Logical Volumes can then be created within a Volume Group.

If you do go the parallel ata (PATA) route, do one drive per port - don't do a drive on master and slave, as ide only allows for one drive to talk at a time on a cable

Since about ATA/33 bus mastering has alleviated some of that concern, but it is still a multi-drop bus and data is written from controller to only one disk at a time, in bus mastering the controller stays relatively 'silent' on the bus and lets the disks talk directly to each other without restriction to performance. Full disk speeds are reached in bus mastering so it's only on 'heavy' IO that one needs to worry. Always use new 80 pin IDE cables and adhere to a bend radius on the cables.

Sniffing the glue holding the Internet together

Mark

1487 posts

Uber Geek

# 203298 25-Mar-2009 17:03

barf:

LVM is the goods. It provides more flexibility than RAID and retains redundancy, definately worth a look since it also facilitates block-device level encryption, snapshots and abstracts the storage model to provide Volume Groups that can span disks of any size (no more wasted bytes to RAID limitations). Logical Volumes can then be created within a Volume Group.

Hi there,

I'm having trouble following what you are saying here ... where is the redundancy in only using Logical Volume Manager ?

LVM on top of a RAID set makes perfect sense, but not just using LVM instead of. They are complimentory tools.
LVM is good for allowing flexible logical volumes, and you can mirror logical volumes, but those LVs are usually just a subset of a Physical Volumes (PV) ,,, you'd have to make really sure of your LV mirroring so ensure you get all you want.
Would be a PITA if you mirrored your MP3 logical volume and then found out the hard way you'd mirrored it to another LV on the same Physical Volume .. you may laugh but it happens a lot! :-) If you LV was on a single RAID-1 PV you could forgo the LV mirror but still benefit from being able to extend LVs (or move them) when you insert extra space later on.

Mark

hads

391 posts

Ultimate Geek

Trusted

# 203337 25-Mar-2009 21:34

Indeed, LVM and RAID are different things.

I personally would use software RAID5 (with LVM if you like), though it does require three disks. These days you can grow a RAID5 easily. Don't be tempted to use fakeRAID either.

I'd go with SATA too, the drives are cheaper and PATA is now old tech. As for drive choice, the Western Digital GP series seem to work well, cool and quiet. That's personal preference though.

nicegear.co.nz - VoIP, DVB and other Linux compatible hardware

meesham

919 posts

Ultimate Geek

Subscriber

# 204825 2-Apr-2009 00:06

I've found the Linux software RAID to be rock solid and would definitely recommend using it when performance isn't a huge issue. Although when you need to grow your array it can take a bit of time, I've just added a 4th 1TB disk to my RAID-5 array, the progress monitor reckons it's going to take 30 hours to grow the array :)