6 drive ZFS replacing drives with bigger capacity

[bobbob1016]

I have a ZFS pool started with 6 1.5TB drives. I now have 4 2TB's and 2 1.5TBs. If I were to replace either one or both of the last two 1.5TB drives with 2TB drives, would I have issues?

My concern is if I add one, the pool might think it's degraded, as it has 5 2TB's.

I'm also unsure if I add two 2TB's if it'll auto-expand to (6*2)-2, or 10TB with 2TB redundant.

Further, if I add two 3 or 4TB's what would it do the (6*2)-2 or is that too much too soon?

 

[danb35]

What, exactly, is your pool configuration? From the math you're doing, it sounds like it's a single RAIDZ1 vdev. If that's the case, yes, when you replace the remaining two disks (one at a time, resilvering after each replacement), the pool capacity will grow. During the replacement process (i.e., while the pool is resilvering during each replacement), the pool will be degraded. This is the pitfall with RAIDZ1 (or RAID 5)--when the extra disk (for redundancy) is gone, any glitch in the data can cause massive problems. But once the resilvering completes, the pool will be healthy.

In any RAIDZ vdev, the capacity is determined by the smallest member disk. Thus, a RAIDZ1 vdev consisting of three 3TB disks and one 1TB disk will have a capacity of (4 - 1) x 1TB ~= 3TB. Right now, you're "wasting" 500 GB of capacity on each of your 2 TB disks, because your vdev is limited by the capacity of your 1.5 TB disks. If you replace those 1.5 TB disks with 3 TB disks, you'd end up "wasting" 1 TB of capacity on each of those, because your vdev would then be limited by the capacity of your 2 TB disks.

 

[SirMaster]

What, exactly, is your pool configuration? From the math you're doing, it sounds like it's a single RAIDZ1 vdev. If that's the case, yes, when you replace the remaining two disks (one at a time, resilvering after each replacement), the pool capacity will grow. During the replacement process (i.e., while the pool is resilvering during each replacement), the pool will be degraded. This is the pitfall with RAIDZ1 (or RAID 5)--when the extra disk (for redundancy) is gone, any glitch in the data can cause massive problems. But once the resilvering completes, the pool will be healthy.

In any RAIDZ vdev, the capacity is determined by the smallest member disk. Thus, a RAIDZ1 vdev consisting of three 3TB disks and one 1TB disk will have a capacity of (4 - 1) x 1TB ~= 3TB. Right now, you're "wasting" 500 GB of capacity on each of your 2 TB disks, because your vdev is limited by the capacity of your 1.5 TB disks. If you replace those 1.5 TB disks with 3 TB disks, you'd end up "wasting" 1 TB of capacity on each of those, because your vdev would then be limited by the capacity of your 2 TB disks.

Does FreeNAS GUI not support online disk replacement?

It's possible in ZFS normally to replace a good disk without bringing the array into a degraded state. Essentially it will resilver to the new disk while keeping the old disk online and connected to the array. Once the resilver is complete it will kick the old drive out automatically.

This is obviously a safer and more reliable way to replace disks especially in a RAIDZ1.

 

[cyberjock]

Does FreeNAS GUI not support online disk replacement?

It's possible in ZFS normally to replace a good disk without bringing the array into a degraded state. Essentially it will resilver to the new disk while keeping the old disk online and connected to the array. Once the resilver is complete it will kick the old drive out automatically.

This is obviously a safer and more reliable way to replace disks especially in a RAIDZ1.

Not at present. It is a planned feature, but no clue when it is to be added. Yes, I agree that this is a pretty lame setback for FreeNAS. But it is what it is.

 

[bobbob1016]

What, exactly, is your pool configuration? From the math you're doing, it sounds like it's a single RAIDZ1 vdev. If that's the case, yes, when you replace the remaining two disks (one at a time, resilvering after each replacement), the pool capacity will grow. During the replacement process (i.e., while the pool is resilvering during each replacement), the pool will be degraded. This is the pitfall with RAIDZ1 (or RAID 5)--when the extra disk (for redundancy) is gone, any glitch in the data can cause massive problems. But once the resilvering completes, the pool will be healthy.

In any RAIDZ vdev, the capacity is determined by the smallest member disk. Thus, a RAIDZ1 vdev consisting of three 3TB disks and one 1TB disk will have a capacity of (4 - 1) x 1TB ~= 3TB. Right now, you're "wasting" 500 GB of capacity on each of your 2 TB disks, because your vdev is limited by the capacity of your 1.5 TB disks. If you replace those 1.5 TB disks with 3 TB disks, you'd end up "wasting" 1 TB of capacity on each of those, because your vdev would then be limited by the capacity of your 2 TB disks.

Ah, ok. Basically, I didn't know if the 5x2TB drives would ignore the 1.5 leftover, and see the system as degraded. Makes sense now, thanks!

 

[Ericloewe]

Not at present. It is a planned feature, but no clue when it is to be added. Yes, I agree that this is a pretty lame setback for FreeNAS. But it is what it is.

Hopefully it'll happen before the block pointer rewrite.