• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Mixing different drives in RAID

Joined
Sep 4, 2008
Messages
10
Hello!

I have some problems with my WD15EADS drive (SMART reports errors) and I want to change it. I don't want to take WD and I want to move to Seagate.

Systems used FreeNAS and ZFS

My questions:

1)Is it okay to mix drives from different manufacturers?
2)Is it okay to mix 5400rpm drives with 7200rpm drives?
3)What about mixing 32MB cache drives with 64MB cache drives?

Thanks!
 
Dell sent me a 500gb RE2 series WD to go with my 3 other RE3 series WD drives. No problems.

At work, I have a RAID 6 on an Adaptec 5805 that has 4 1tb Seagate ES.2 drives and 4 WD RE3 1TB drives. No problems.

You're probably ok.
 
It will be fine on linux or freebsd software raid. The smallest drive will be the maximum raid unit. However performance will be nonoptimal with mixed drives.
 
I am not looking for performance. I am looking for stability of the RAID. I will use FreeNAS with ZFS and I think it will be fine, but wanted to know if there are any horror stories about mixing different drives.

Thank You for your answers!
 
1) Yes, but you'll want to have the same size drives, and you should try to pick a similarly performing drive or else you will really kill performance.

2) Never tried it, but mixing drives of differing rpm in RAID sounds like a bad idea which wouldn't have any benefits.

3) Shouldn't be a problem.

If you're looking for maximum reliability instead of performance, mixing drives from different manufacturers and manufacture dates is a very good idea. It's a huge risk having a large number of drives of the same model and manufacture date in a RAID array, since a manufacturer defect or firmware bug could cause all of your drives to fail at the exact same time or in close proximity to each other. Just ask someone who went through the IBM Deathstar catastrophe.
 
You should be fine if you're going to use ZFS. You won't have alignment issues and not too low stripesize, since ZFS has dynamic stripe size that adapts to the file size. Only possible if the Filesystem and RAID-engine are one package.
 
Thanks for info. I am testing now 2 drives with ZFS, both have 7200rpm but one has 8MB cache and the other 16MB cache to see how it works.

I will keep it like that for a week and if it is okay I will replace the 16MB cache drive with a 5400rpm drive with 32MB cache and test it for another week. If all goes well I will start to put the data on the NAS.
 
Please note that mixing 5400rpm and 7200rpm drives has far more effect than mixing drives with different cache sizes. The cache size is not that important.
 
NECRO'D

Don't really want to revive an old thread for my first post, but its better than creating a new one IMO.

Here's my question of the same vain as the OP. I currently have two WD Caviar Green 1TB drives (WD10EADS so 5400rpm, 32mb cache with the WDTLER modified) and two Samsung Spinpoint F2 1TB drives (HD103SI also 5400rmp, 32mb cache). Would I be able to safely/reliably combine these 4 drives into a single RAID array? Also, I'm currently just using my mobo's onboard RAID. Is a benefit of using a dedicated RAID controller card the fact that you can easily swap the card + drives into a new machine without losing your data?

Sorry for the n00bish questions. I've been running RAID arrays for a couple years now in my personal PCs but never really learned all that much about it...
 
NECRO'D

Don't really want to revive an old thread for my first post, but its better than creating a new one IMO.

Here's my question of the same vain as the OP. I currently have two WD Caviar Green 1TB drives (WD10EADS so 5400rpm, 32mb cache with the WDTLER modified) and two Samsung Spinpoint F2 1TB drives (HD103SI also 5400rmp, 32mb cache). Would I be able to safely/reliably combine these 4 drives into a single RAID array? Also, I'm currently just using my mobo's onboard RAID. Is a benefit of using a dedicated RAID controller card the fact that you can easily swap the card + drives into a new machine without losing your data?

Sorry for the n00bish questions. I've been running RAID arrays for a couple years now in my personal PCs but never really learned all that much about it...

Bump for response. I've been reading conflicting opinions on various sites when trying to search, so any feedback would be appreciated! I'm thinking this machine will eventually migrate from a Win7 box sharing drives to a NAS, but need to get this RAID stuff sorted out before then.
 
You can mix drives, but performance-wise you might take a small hit when mixing drives with different performance characteristics.

I would highly advise maintaining a backup of your data; that will be your best protection against loss of data. For example, consider creating two RAID arrays each consisting of two of your samsung/WD HDDs, using one array as backup.

The most reliable way of storing files may be using ZFS, but consider at least a backup to protect your data.
 
If you are looking for stability then I'd say don't mix drives. I can almost tell you are going to have a random 'drop' in drives down the line due to the difference in drive latency.

The easiest way for you to tell is to just create the RAID with the drives you want with multiple volume and heavily stress the RAID with IOmeter with 16 workers and small stripe size/IO for 3-7 days on each volume and see if the controller event log report timeout error. (At least on the Areca controller)
 
If you are looking for stability then I'd say don't mix drives. I can almost tell you are going to have a random 'drop' in drives down the line due to the difference in drive latency.

The easiest way for you to tell is to just create the RAID with the drives you want with multiple volume and heavily stress the RAID with IOmeter with 16 workers and small stripe size/IO for 3-7 days on each volume and see if the controller event log report timeout error. (At least on the Areca controller)

Ok. I guess I was just hoping since they are both similarly spec'd 1TB drives it wouldn't cause much issue. Guess I'll keep my plans of migrating the PC to more of a NAS on the back burner until I can buy up a batch of matching high-capacity drives. For now I will need to pick up a RAID controller card though....
 
Since you mention NAS do you really need a raid controller? Most times you just use the motherboard ports since they are generally fast and will not be using the raid functionality of the controller under most NAS operating systems (linux, freebsd, opensolaris ...) that have good OS raid.
 
Since you mention NAS do you really need a raid controller? Most times you just use the motherboard ports since they are generally fast and will not be using the raid functionality of the controller under most NAS operating systems (linux, freebsd, opensolaris ...) that have good OS raid.

For the meantime I will need one, as I don't want to rely on my onboard RAID like I am now as I have had it drop the array before for no good reason (which is why I bought the couple of Samsung drives). I have heard plenty of talk reading around here about using the software raid/fake raid with a NAS, but yet sites for software such as FreeNAS mention support for controller cards, so I'm unsure. My main thing is being able to take the card + drives from one PC to another if I need to upgrade the hardware, but then again, I dont think there would be a huge need to upgrade hardware when it comes to a NAS, other than for failures.
 
I am not saying fakeraid. The raid I am talking about is provided by the OS on these systems and in this case whatever SATA controller you have it is in normal SATA mode. The OS raid is totally hardware independent. Just grab the disks put them in some other motherboard and install the os (or even boot a live cd/usb stick) and the raid array will function.
 
@LiNxGT: This may be of interest to you:
http://hardforum.com/showpost.php?p=1036118140&postcount=7

If you use a FakeRAID controller like Promise/Silicon Image/JMicron on Linux/FreeBSD then you can use it as normal SATA controller; as that's all what the hardware is really. It relies on Windows-drivers to do the RAID part. In Linux/FreeBSD, you simply use the OS-supplied Software RAID (which can be ZFS).

Thus, no drive dropouts even without TLER. Note that you still will encounter drive dropouts even with some high-end hardware RAID controllers. Only software-RAID on Linux/BSD would escape the TLER-tax.
 
@LiNxGT: This may be of interest to you:
http://hardforum.com/showpost.php?p=1036118140&postcount=7

If you use a FakeRAID controller like Promise/Silicon Image/JMicron on Linux/FreeBSD then you can use it as normal SATA controller; as that's all what the hardware is really. It relies on Windows-drivers to do the RAID part. In Linux/FreeBSD, you simply use the OS-supplied Software RAID (which can be ZFS).

Thus, no drive dropouts even without TLER. Note that you still will encounter drive dropouts even with some high-end hardware RAID controllers. Only software-RAID on Linux/BSD would escape the TLER-tax.

Wow, thanks for that post sub.mesa. Bear with me hear as I'll admit again that I am really new to this stuff, even Linux in general. There has been mention of using the OS software RAID, would it still be recommended to avoid mixing my WD and Samsung drives in an array this way? I'm thinking I may pick up a couple more of the Samsung drives since they're ~$60 (even though they seem to be discontinued). Are there certain distributions of Linux that I'd have to look at to run this sort of configuration? I've only briefly played around with Ubuntu, and I've been running Mint on a separate drive in my laptop, but I'm not sure if these more "home" version of Linux would have the RAID functionality I'm looking for? Also, I've read a bit about ZFS, but I guess I want to be sure about how Windows/Apple machines would handle reading/writing to these network drives, if they can at all. I don't currently own an Apple device but I can't rule out the possibility down the line. I'd also probably need a torrent client with a web interface (I currently use uTorrent and its WebUI) and something like PS3 Media Server... Maybe I should have created my own thread after all =)
 
If you use a FakeRAID controller like Promise/Silicon Image/JMicron on Linux/FreeBSD then you can use it as normal SATA controller; as that's all what the hardware is really. It relies on Windows-drivers to do the RAID part. In Linux/FreeBSD, you simply use the OS-supplied Software RAID (which can be ZFS).

sub would you mind taking a look at THIS post please.
 
Wow, thanks for that post sub.mesa. Bear with me hear as I'll admit again that I am really new to this stuff, even Linux in general. There has been mention of using the OS software RAID, would it still be recommended to avoid mixing my WD and Samsung drives in an array this way?
The mixing of HDDs would affect performance a bit, but it should not change the way your RAID works or operates; it's not like your RAID controller treats the other disks differently; those disk just have slightly different performance characteristics which can be amplified. In some cases, this can cause the slower disks to slow down faster disks in the same array.

Aside from this, i can see no reason why to avoid two reasonably similar disks of different make in a Linux software array. That should 'just work'.

Also, I've read a bit about ZFS, but I guess I want to be sure about how Windows/Apple machines would handle reading/writing to these network drives
The same as if you used Linux and let your Apple machines store on Ext4fs, XFS, JFS, ReiserFS or any other filesystem. When using network filesharing, you can distinguish two protocols:

NAS-protocol (SMB, CIFS, NFS, FTP) - the server controls the filesystem; clients don't even know what filesystem they are on.
SAN-protocol (iSCSI, ATA-over-Ethernet) - the client controls the filesystem; the server doesn't know what kind of files the client stores. The server just sees a virtual harddisk of 5GB for example; and it cannot be used for more than one client.

Most common is NAS-protocol SMB/CIFS; or windows filesharing. Your mac supports this too, and likely also the faster NFS protocol. In both of these protocols, your mac client does not control the remote filesystem. The result: you can use whatever backend you wish. ZFS? Sure; Mac OSX/Windows won't even know they are storing on ZFS.

If this didn't work then ZFS would be useless to home NAS people. Since it does work, it allows you to combine the world of legacy I/O with the world of next-generation checksumming filesystems, and thus combine the best of both worlds.

Concluding: it is perfectly fine to let Windows/Mac store files on ZFS filesystem over the network; many people use ZFS this way.
 
Aside from this, i can see no reason why to avoid two reasonably similar disks of different make in a Linux software array. That should 'just work'.


The same as if you used Linux ...

Thanks again sub.mesa. So based upon everything thats been discussed, would a more pedestrian distribution of Linux (such as Ubuntu or Mint) be able to accomplish functioning as a NAS utilizing the NAS-protocol and have software RAID with ZFS, or would I have to look into something else? As I mentioned before, the only real software functionality I'd be looking for on top of the NAS/RAID would be a torrent client with a web interface and PS3 Media Server or something like it that can properly transcode with subtitles.
 
would a more pedestrian distribution of Linux (such as Ubuntu or Mint) be able to accomplish functioning as a NAS utilizing the NAS-protocol and have software RAID with ZFS

ZFS is more complicated in linux because it is not in the kernel. There is a port which includes a fuse module but I am not sure that should be used in production environments and in any case it will require you to understand what you are doing. I would not do that if I was not highly familiar with linux.

ZFS is however available on freebsd and opensolaris and several other options. There are many threads about this. sub.mesa even has contributed a GUI to help administer ZFS under freebsd.
 
Well i assume he is considering either Linux software RAID with conventional filesystem (XFS, JFS, Ext4) OR FreeNAS+ZFS; not Linux+ZFS as that's not mature at all.

If you choose ZFS, then indeed you may be able to run my Mesa ZFS-interface in the future; you can even test it right now inside Virtualbox VM to review its potential. But you should not use it for real yet. Instead, using OpenSolaris/FreeNAS/FreeBSD now for ZFS and migrating later to my Mesa project would be possible.
 
Well i assume he is considering either Linux software RAID with conventional filesystem (XFS, JFS, Ext4) OR FreeNAS+ZFS; not Linux+ZFS as that's not mature at all.

If you choose ZFS, then indeed you may be able to run my Mesa ZFS-interface in the future; you can even test it right now inside Virtualbox VM to review its potential. But you should not use it for real yet. Instead, using OpenSolaris/FreeNAS/FreeBSD now for ZFS and migrating later to my Mesa project would be possible.

So it seems from my (bit of) research, you would get FreeNAS (or whatever OS) up and running, and then add in your storage drives, formatting them to ZFS, and then you configure them using zpool? This whole setup should still be considered the same as typical RAID in that the disks should match, correct?
 
Hmm you probably still have some questions about how ZFS works.

Think of ZFS as RAID engine (ZPOOL) and filesystem (ZFS). The two act more or less independently:

zpool = one pool of three raid-z arrays for example
zfs = the filesystem that will be storing files

So the zpool part could be considered the RAID part and the zfs part the normal filesystem. Zpools can be made from any combination, a weird example:

pool "bigpool" consisting of:

single disk
- ad0
mirror
- ad1
- ad2
mirror
- ad3
- ad4
raidz
- ad5
- ad6
- ad7

This pool contains a single disk without redundancy, two mirrors and a raid-z all in one pool. The ZFS filesystem then stores information on this pool; it doesn't matter that some vdevs (= mirror or raidz or single disk) are bigger than others.

So you can mix different drives, even of different capacity. You can't really mix drives of different capacity inside the same vdev (i.e. a mirror needs disks of the same size) but you can have mirror1 be bigger than mirror2 if they are both in the same pool. In other words, the vdevs themselves can differ in size.

So you could have a really strange complex setup of all kinds of disks, and still have it bootable.

The only thing that limits hardware usability right now is EARS disks on RAID-Z; which perform much less well than they are capable of. EARS disks in mirror configuration works alot better. This has to do with variable stripesizes causing unaligned end offsets in I/O done to EARS disks in RAID-Z vdevs. Shortly said: avoid EARS for now if you can; if you can't assign them to mirrors instead of raid-z.

Feel free to ask other stuff; only when you feel you understand it yourself should you stop asking questions. ;-)

But take one step ahead; ZFS brings alot of theory to you, but it's good to actually know what that theory means in reality. So trying out things likely enriches your knowledge about ZFS features.
 
So the zpool part could be considered the RAID part and the zfs part the normal filesystem. Zpools can be made from any combination.

So essentially zpool is the "RAID" aspect of ZFS, but in the grand scheme of things the members of the pool don't have to be in any form of RAID (standalone)? Taking your example of the non-redundant disk, the 2 separate mirrors, and the raidz array, would the single disk pose as the OS disk, or would you maintain the OS disk outside of the pool? When it comes to the NAS aspect, would the client-side machines see each vdev as an individual drive location, or would it see the pool as a drive with the vdevs as folders or something else entirely? You said to keep asking questions if I have them, hope you're prepared for the potential can-of-worms that could be :)
 
Back
Top