I have created a 3-drive Linux RAID5 (md) array out of brand new Seagate Barracuda drives. I can measure the performance of the whole array by using: ioping -s 8k -c 10 -D -W .
, where the current directory is backed by a RAID device (/dev/md1
). Is there any way I can measure the IO performance (not only read, but also write) off the individual drives, e.g., /dev/sda1
, etc, without damaging the filesystem?
Asked
Active
Viewed 533 times
0

sherlock
- 141
- 5
-
1Please do not use RAID5 if data is of any importance to you. – Tero Kilkanen Jul 20 '21 at 06:32
-
I agree with @TeroKilkanen, a three drive RAID5 has the capacity of 2x one drive. A three drive RAID1E has 1.5x the capacity of one drive. With all the downsides to RAID5, It hardly seems worth it. But, a half-answer to your question: you can benchmark with full stripe writes. If you do that, there shouldn't be any reading going on. Other than that, there aren't any good options. You can fail a drive out and do your benchmarking on it. Wipe the drive before putting it back in to make absolutely sure you don't recover from the bitmap. – Mike Andrews Jul 30 '21 at 18:45