Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Is it recommended to use software RAID-1 in a production environment? Due to budget constraints, I cannot get a hardware RAID controller for a new server.

While I'm aware of a slight performance decrease, is it really that much of a difference?

share|improve this question

7 Answers 7

up vote 5 down vote accepted

Actually, you may be happier with software RAID as it is more flexible in regards to disk sizes and types. I would almost recommend it over hardware RAID because of this and what ptman says about having to have backup controller in case it dies.

PS I wouldn't use RAID5 hardware or software. http://www.baarf.com/ I'm not just accepting what they say. I've been disappointed by RAID5 performance and found the answer why. Do a RAID 10 or similar.

share|improve this answer
    
If you are using Linux then the new fangled RAID10 driver (as opposed to layering RAID0 over RAID1) can work over three drives like RAID5 (an arrangement like that called RAID1E by some controllers). In this mode you get the same redundancy as RAID5 (one drive can die and the array survives) but without the "read plus parity calc plus two writes per write" performance kick. You should see write performance akin to a two drive RAID1 and read performance similar to a two drive RAID0. The newer offset option promises to boost the read performance further, but I've not experimented with that yet. –  David Spillett Mar 29 '10 at 9:14

I use the FreeBSD software RAID-1 (gmirror) on production web servers, and the performance impact isn't noticeable.

I wouldn't suggest Software RAID-5 because of performance overhead (the parity calculation is best left to a dedicated RAID controller). and I wouldn't run it on a database server in production, but if disk I/O isn't going to be too heavy it's probably fine.

share|improve this answer

Linux software RAID1 is fine. And in some ways better than hardware RAID. I hated not being able to upgrade the kernel of a RHEL installation because the binary drivers for the hardware RAID weren't updated for the newer kernel. And if the RAID card dies you need to get another one to get at the data (well, not necessarily with RAID1, but with RAID5 you would), but with software RAID any machine will do.

share|improve this answer

Linux software raid1 is running fine on my production environment. 2 x Seagate Barracuda 7200rpm 16mb 500GB Hard Drives:

/dev/md0:

Timing cached reads: 1894 MB in 2.00 seconds = 946.49 MB/sec

Timing buffered disk reads: 312 MB in 3.01 seconds = 103.62 MB/sec

share|improve this answer

I used to use hardware RAID1 with Linux… and I have replaced all of them with mdadm. I found the hardware solutions (not a top-end, I guess, but still not the cheapest 'pseudo-raid') quite unreliable. Even if it worked for some time after a kernel upgrade it would stop working correctly. It seemed that finding a right firmware+driver combination could help, but it was easier to set up software RAID instead.

In other case, the RAID controller simply failed. And it was hard to get the data back from the disks without it. With Linux RAID1 you can always access your data if it is still on a disk, you don't depend on the specific piece of hardware… and the performance is not much lower.

As I use RAID for reliability, software solution seem just better for me. I found it more reliable.

share|improve this answer

Yes! Considering you can't aquire a hardware raid setup, raid1 done is software is way better than no raid at all.

share|improve this answer

I'm basically in agreement with what's already been said, Linux software RAID works great.

I wanted to add a note of caution, however: I've had uneven luck with configurations where the boot and/or root drives are mirrored. Although data integrity is retained, two operational issues have come up.

  1. The first varies wildly by hardware and kernel versions: I've had boxes that locked up or even panicked when a drive died in certain ways. Newer hardware with recent kernels seems to have vastly improved this situation, but I can't guarantee it's gone away.

    • I have seen this with hardware controllers in one case, too. Older Dell 1850s had a bug that often caused their PERC cards to drop off the face of the earth when a drive died until a hard reset was performed.
  2. The second is getting a machine with a bad "boot" drive to boot. In most cases, grub or other applicable bootloader will not be automatically installed on both disks, so if the primary boot drive goes and the server gets rebooted, it won't come back up without manual intervention. This can be resolved, of course, by carefully ensuring grub is properly installed on both drives.

    • Another failure mode can occur if the drive fails in such a way that the BIOS tries to boot from it, but the drive cannot fully complete the process.
share|improve this answer
    
thanks, good to know. –  Gareth Mar 30 '10 at 10:01

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.