[SATLUG] OT: RAID drives
brad at shub-internet.org
Wed May 11 19:20:48 CDT 2011
On May 11, 2011, at 5:48 PM, Alan Lesmerises wrote:
> I have a general question about RAIDs that I haven't seen an answer to .... yet.
> How closely do the drives need to match in a RAID, when you're looking for replacements for drives that are in an existing RAID?
It depends on the type of RAID, and how it's implemented. For something like a ZFS "zspool" or a Drobo "BeyondRAID" method, they should be able to make use of whatever you've got.
For a traditional RAID, the mechanisms should be identical make and model, and sometimes even that is not enough. You can probably get something that will kinda-semi-sorta-mostly work if they're "close enough", but performance is likely to suck -- and this can be a problem for you even if they are supposedly the same make and model, because sometimes there are internal variations that are enough to make the raidset perform very badly, even when the mechanisms are supposedly identical. Or sometimes manufacturers will re-use the same model name and/or number on a totally different mechanism, and you don't find out until they show up on your doorstep what it is that you've really purchased.
> Specifically, I have several identical drives in a RAID 5, but only 1 spare right now. In searching for sources for additional spare drives, the specific model drive I currently have doesn't seem to be readily available (not unless you're willing to spend huge amounts of $). But I can find some drives that are pretty close to the ones I already have. So the question is: How close is close enough?
How much do you care about performance?
> - How close does the new drive need to be (in terms of capacity) compared to the drive it's replacing to keep from totally screwing-up the RAID?
In cases where you simply don't have a choice, going with newer drives that are larger in capacity, but otherwise the same make, model, speed, etc... will be likely to get you a raidset that won't degrade too badly, and may not hurt your performance at all. But the further you get away from being identical, the more you increase your chances that something really perverse will happen and the whole thing will go down the crapper.
I know that some vendors (like NetApp) will replace older lower-capacity mechanisms with newer higher-capacity mechanisms of the same make and model when there are no more of the older mechanisms to be had, but they do go through an extensive qualification process before they allow any drives at all into their units. You probably won't have that kind of luxury, so you won't know for sure if the same methodology will work for you, until you've made the decision and committed yourself.
> - Do they need to be the same speed (7200 vs 10k or 15k), or does that make much difference (in terms of data integrity)? I do recognize that one slow drive could create a bottleneck, at least under _some_ circumstances.
I don't think difference in speed is likely to impact data integrity, but if the new drives are smaller in capacity than the old ones, that would definitely be a problem. And major differences like rotational speed would probably keep the RAID software from allowing you to add the new mechanisms to the old raidset.
> - Are there any other considerations I should know about?
> What do you think?
The real answer is "it depends". You almost certainly won't know until after you've made your decision and committed yourself, and by then it will be too late.
If you do manage to buy new mechanisms that are incompatible with the old ones, hopefully the RAID software will be smart enough to prevent you from adding them to the raidset. In that case, the worst outcome that you would be likely to have would be that performance would be so badly impacted that the filesystem would be effectively unusable, even though it would technically still be in existence.
OTOH, it is entirely possible that the new mechanisms might not cause any problems for you at all, and everything could go down without any rough spots of any kind.
What you would probably be most concerned about is when things seem to go okay, but you find out later that there were silent errors that cropped up which resulted in unrecoverable corruption of the filesystem. In theory, the RAID software should hopefully prevent this from occurring, but it's hard to be 100% certain about that.
Personally, in cases like this, I tend to err on the side of caution, and try to make the new mechanisms as identical as possible to the old ones -- same make & model, same rotational speed, same revision of the drive controller firmware, same everything.
If that's not possible, then I'll go with the closest match that I can get, so long as the new drive is at least as fast as the old one and has at least as much raw & formatted capacity as the old one, and as similar as possible in terms of make & model.
At the end of the day, you pays your money, and youse takes your chances. And you keep an eye on the dice at all times.
Brad Knowles <brad at shub-internet.org>
LinkedIn Profile: <http://tinyurl.com/y8kpxu>
More information about the SATLUG