Adding new vdev to existing pool vs. create new pool

[idgar]

Hi,
if one have a current pool that includes 1 vdev in RaidZ1 (3 disks) and want to add a new RaidZ1 vdev (3 disks), what is the pros and cons of:
1. creating a new pool vs. extending the existing pool.
2. in case of disk faults, which will be easier to handle? (if for example it needs at worst case data recovery).
3. upgrading vdev in the future to larger disks.

All I know that it is easier to know which dataset belongs to which vdev when there are 2 pools, but I'm not sure how a pool with multiple vdevs spreads its data/datasets.

 

[Tony-1971]

Hello,

I try to answer.
1. If you create a new pool with 6 disk, then this operation destroy your data. If the old pool remain and you add a new pool then the size of old pool is not growing (adding a vdev can be done if there is low space on pool).
2. If I undestand your question, probably there is no difference.
3. Also I don't think there are difference.

Best Regards,
Antonio

 

[idgar]

Thanks @Tony-1971 for the answer.
I'm still missing some details.
I saw in the post is it possible to remove a vdev from a pool that it is not possible to remove a vdev from a pool and if I understand correctly, once you've added a vdev it gets stuck there forever (and with the same setup, e.g. RaidZ1 of 3 disks, not sure about the size of vdev).

so if you cannot remove a vdev, it means that you are unable to:
1. get rid of that vdev in case you have enough free space in your pool and you want to save money as you don't need this space currently.
2. replace that vdev with another vdev setup when you buy new disks. let's say you have RaidZ1 with 3 disks today of 1TB each and you want to replace it with new RaidZ1 of 4 disks each 2TB. it sounds like it is only possible if that vdev configured alone in a pool and the new vdev is created in a new pool. only then it's possible to migrate.

it's sounds to me that the cons of 1 pool with multi vdev are:
1. ease of use as everything is in one pool and you don't need to think about where to position your dataset.
2. higher chance to utilize your disk space in case you have large data.
3. create dataset that is larger a vdev limit.
4. performance, as a dataset can potentially spread across multiple vdevs.

I wonder if what I said true. I will be glad to hear some opinions.

Thanks in advance!

 

[Tony-1971]

You can replace disks in a vdev (if there is enough redundancy) with bigger disk, so if you have RaidZ1 with 3 1TB disk you can replace them with 3 2TB disks (one disk at a time, so you have to wait long time).
At this time is not possible to expand a vdev only adding one disk, so if you have RaidZ1 with 3 1TB disk you can't add another 1TB disk to obtain a bigger vdev.

Best Regards,
Antonio