AVB
Contributor
- Joined
- Apr 29, 2012
- Messages
- 174
I'm working on 2 Vdev raid z2 pool with 8 eight TB and 8 four TB drives. I'm doing an upgrade to the 4TB drives to 6 TB drives and it is taking far longer than I would have thought - in the neighbor hood of 65 hours per drive to resilver the 3 I have done. The pool before the upgrade was at 61% so total capacity shouldn't be the problem. When I ran iostat -v it showed that the data per drive was just about equal at 3.16TB each in both vdevs. While 3.16TB is only about 43% capacity of the 8TB (really 7.21TB) drives it is 87% of the 4TB (3.64TB) drives. I'm thinking that the 4TB drives being so full is what is really slowing things down,
So the question is: Does rebalancing happen only across the number of Vdevs in the pool or can it be set to balance according to Vdev capaciity? In my case the 8TB Vdev would get 2/3 rds of the data and the 4TB Vdev 1/3rd. If balancing is only based on the number of Vdevs then that is a pretty good argument to have all the Vdevs exactly the same since I assume having one Vdev "full" slows everything down.
Thoughts, comments and opinions welcome.
So the question is: Does rebalancing happen only across the number of Vdevs in the pool or can it be set to balance according to Vdev capaciity? In my case the 8TB Vdev would get 2/3 rds of the data and the 4TB Vdev 1/3rd. If balancing is only based on the number of Vdevs then that is a pretty good argument to have all the Vdevs exactly the same since I assume having one Vdev "full" slows everything down.
Thoughts, comments and opinions welcome.
