RAID5 Question

martinw79

n00b
Joined
Nov 16, 2007
Messages
19
If I have a RAID5 volume with only 3 drives, what happens when one drive fails? Does the system stay up and running with decreased performance or am I down until I replace a drive?

I was under the impression everything stays online but the performance takes a hit, the server is NOT down until I replace the drive.

I'm asking because an IT consultant just told us we would be down if 1 of the 3 drives in a RAID5 crashes.

I know the more drives in RAID5 the better, but we have a budget for storage of these SAS drives.
 
If one of the drives dies, then you're still up and running, but at a slower rate than before both because you have less parity to read from and (assuming you have a hot spare) the overhead of rebuilding the array.

More drives in a RAID 5 array is actually worse from a reliability standpoint. Not only do you have more points of failure, but the amount of time it takes to rebuild an array due to a dropped disk becomes substantial. With large capacity drives, there is no way a 16x2TB RAID 5 array would fly - the amount of time it would take to rebuild it would be asking for trouble.
 
I am not so sure about decreased performance well at least I have not seen this on linux software raid. With that said I would not have my data running as raid 0 for long..
 
vsboxerboy;

If we have RAID5 with 3 drives and no spares / no online spare. All 3 volumes make up the RAID5 obviously. If 1 drive fails are we still up? From what I gathered the online spare simply lets the rebuild process start on it own w/o end user interaction. Am I correct?
 
The array should stay up at nearly the same performance but no redundancy. There is a little overhead in using the parity on each read. At least this is what I have seen on the dozens of raid 5 arrays I have here at work. HW controllers could act differently in this situation. You will have noticeable reduced performance when you do a rebuild.
 
raid 5 gives you a 1 drive tolerance, 1 drive dies, you stay up, 2 drives die, your screwed.

hotspare means as you said. the system will automatically start rebuilding the system with the failed one

Remember Raid is not a backup, but simply for redundancy, you should have up to date backups as well.

You will have performance loss, period, you are working with 2 drives vs 3 drives.
 
With large capacity drives, there is no way a 16x2TB RAID 5 array would fly - the amount of time it would take to rebuild it would be asking for trouble.

Acceptable rebuild times are quite subjective.
I have a 9x 2TB RAID6 array and i forced a rebuild and it only took 22 hours
 
yes, but you have raid 6, not raid 5, and you have 9 drives vs 16, as said more drives = more points if failure, raid 6 allows for 2 failres, raid 5 one, in that time frame of 22 hours, which is a long time another drive can die and your screwed BUT if you have this kind of setup with important data, backups should be available.
 
You will have performance loss, period, you are working with 2 drives vs 3 drives.

When you have 3 drives in a 3 drive raid5 for every block 2 drives are being used and the other is ignored. Of course the parity is distributed so the next block will use the drives used will change. Yes there is a performance loss but no it's not like 50% or anything.
 
I have a 9x 2TB RAID6 array and i forced a rebuild and it only took 22 hours

I call that slow. Linux software raid is at least 2 times this speed with modern cpus.
 
When you have 3 drives in a 3 drive raid5 for every block 2 drives are being used and the other is ignored. Of course the parity is distributed so the next block will use the drives used will change. Yes there is a performance loss but no it's not like 50% or anything.

yea definitly not 50%. last time i had a drive fail on my 2TB Raid5 (3 x 1TB Drives) i noticed about a 20% or so decrease in performance.

I didn't have a hot spare and i didnt have the cash to buy another drive right away. Went about two weeks with it like that. I was sweating bullets the whole time.
 
With large capacity drives, there is no way a 16x2TB RAID 5 array would fly - the amount of time it would take to rebuild it would be asking for trouble.
Acceptable rebuild times are quite subjective.
I have a 9x 2TB RAID6 array and i forced a rebuild and it only took 22 hours

I think he was referring less to the time involved and more to the lack of protection for all those drives during that time.

Since all the other drives need to be accessed to rebuild the failed drive, not only are you running without protection but you're also putting extra load on the remaining drives. This can actually cause additional drive failures before you've finished the rebuild and are protected again.

Using these two examples, the RAID5 array would have to work with 15 drives' worth of data with no protection, while your RAID6 would have to work with 7 drives' worth of data with 1 disk for protection. I consider that a huge difference.
 
Back
Top