Possible disk wonkiness on raid

[Joose]

I've got my Steam install on a couple of old 500gig drives pushed together with raid to make a 1Tb drive, and I think one of them might be failing. Every now and then a game will throw a wobbler, and it will turn out that some file or other has become corrupted and needs redownloading. This isn't a massive problem, as the only thing on these drives is Steam, and I have another drive ready to replace the both of them if needs be, so if the worst happens I will just have to redownload my games (praise be to 50meg internets). It seems unlikely to me that both drives will be failing at the same time though, so is there any easy way to tell if the problem *is* a drive about to fall over, and if it is, which drive is the problem one? If its just the one, I can put the working 500gig to use elsewhere.

[deject]

I'm not sure how the RAID will complicate things, but in general, the best way to determine the health of a hard drive is to use the manufacturer's diagnostic tools to scan the drive.

Western Digital: http://support.wdc.com/product/download ... =3&lang=en
Seagate: http://www.seagate.com/ww/v/index.jsp?l ... 04090aRCRD
Hitachi: http://www.hitachigst.com/support/downloads/#DFT

[HereComesPete]

One of the easiest ways to tell if you have a failing drive is exactly what you've got there. Disappearing data. A scandisk run can also point out if there's problems. Most manufacturers also produce diagnostic programs that can tell you if something is on the fritz.

Editz - bums, beaten by the deej.

[Joose]

Well, I tried the WD one, but it fails on both drives almost instantly. Googling suggests that the tool wont work on raid, which is a pain.

Im just going to leave it and see what happens. If the drive(s) cack themselves the only thing ive lost is steam, which is easily replacable.
Thanks anyway!

[FatherJack]

I use symbolic links to distribute my Steam games across multiple drives - one fast one and one big one, GameSave Manager creates and copies the data with its Steam Spreader function, but the functionality is actually part of Windows, so programs don't know any different and report 37GB free on a 300GB disk that a dir /s shows as having 740GB on it. I use the same method to redirect save data for games that insist on dumping a ton of data on the C: drive, like The Witcher and Sins of a Solar Empire.

I used to use striped RAID0, which while giving good performance produced far too many errors. The failure rate is theorectically doubled because either disk failing means the whole array fails, but in practice I've found it much higher than that - as both disks are in near-constant use their individual failure rate is increased, and any power-cuts or non-clean shutdowns increase the failure rate of the array. Also any defrag/disk-checking tasks (which don't really make any sense on an array) cause very excessive activity on them. The expense of buying replacement identical 300GB VelociRaptors became prohibitive in the end, especially with the performance of Samsung Spinpoints catching them up (apart from in seek times) eventually.

If you're only using a JBOD array you're getting all those negatives without the performance gain, so I'd suggest you backup, break the array and test each disk individually. They may both pass meaning it's only a peculiarity of RAID that's causing the errors, so you could happily carry on using them as two seperate disks.

HDTune is a great piece of software for scanning and speed-testing disks, but it won't tell you which disk within an array is dodgy, only manufacturer tools will do that - by which I mean the manufacturer of your motherboard - the RAID controller. There are usually a few you can get at by pressing a key at boot-up, but manufacturers like nVidia also do some in-Windows ones too. Most often you have to break the array to find out, though.