0
I have a few decent, but differently-sized SSDs lying around that I would like to use in a stripe in Freenas. The sizes are: 2 x 240GB, 1 x 256GB, 1 x 1TB.
Using a stripe, this gets me to ~1.73TB. Now Freenas warns that differently sized disks are not recommended in a VDEV; however, this warning seems to have its roots in avoiding this scenario when using Raid-Z, as one would obviously lose space.
My intentions for this pool would be:
- a 500GB ZVOL shared over iSCSI for an ESXi running up to 10 VMs, with daily backed up snapshots (copy the daily snapshot to a different pool)
- use the rest of the space as fast throw-away space (i.e. the data there is not backed up and if it's lost, then nobody cares)
So my question is: What are the potential implications I am looking at? What will the performance be (IOPS and throughput esp. for the VMs)? Any reason I should not do this?
Thanks!
user654123
Posted 2019-08-17T10:41:52.987
Reputation: 213
With no redundancy and if you're not shutting down all the VMs while taking a snapshot, the first thing I'd research is whether or not a filesystem snapshot of a running VM is a guaranteed-to-be-viable backup strategy. Offhand I don't know if it is or isn't, but if it's not and you have a failure that results in corrupt VMs, you will likely have irretrievable data loss. Because with a stripe setup like that, any one disk failure is likely to de facto drop the entire pool anyway as large files will likely have blocks over all the disks. – Andrew Henle – 2019-08-18T15:56:13.347
Thanks for your comment! There is a feature in Freenas that snapshots all VMs in ESXi, before the ZVOL snapshot is made, that should leave them in a consistent state. You are right about the one-disk failure, but the daily backups and the other throw-away data should have me covered. What do you think about the performance of such a setup? How does such a stripe behave? Can the IOPS and throughput of all disks together be exploited? – user654123 – 2019-08-21T09:18:27.160