This is because the 2xVDEVs are forming a pool. The pool consists of one or more VDEVs, right?.
1 Like
Pools are made up of one or more vdevs. Vdevs are made from one or more physical devices. Fault tolerance is configured at the vdev level be it mirrors or RAID Z1/2/3. However the rule is true in all of these configurations if you lose one or more of your vdevs then you lose your pool. If you wish you could create two pools each with their own vdevs and in that case they would be completely separate from a fault tolerance perspective but also an available space perspective.
1 Like
No, RAID0 is super redundant, right? 
So if I have 2 VDEVs, both with RAIDZ2, the datasets are striped across the 2 VDEVs? (You don’t put datasets on one VDEV or the other). If so, then to big win for doing 2 RAIDZ2 vdevs in this instance is only to save on resilver time (and possible impact to other drives while resilvering)? Additionally, it will perform better?
If you had really different data storage needs, would you create 2 pools each with a RAIDZ2 instead (since you would have different data sets)?
Yes, that is what I’m saying. 3 disks fail in vdev1 the entire pool fails. You could have 2 disks fail in vdev1 and 2 disks fail in vdev2 and the pool will still be fine, but in short: if you lose a vdev, your entire pool fails.
Of course, you can make each vdev into its own pools and failure of 1 vdev will only affect that pool, but that is not a common use case.
1 Like
Totally makes sense.
The beauty of ZFS and TrueNAS is the ability to pool (forgive the pun) storage and grow as you need. Although having two different pools for different workloads is fine it does feel to me inherently going against the general idea. More vdevs also means more performance so often separating vdevs into different pools limits your overall performance depending on your workload. I’ll be honest I haven’t read the entire thread but if you haven’t already it’s best to explain your workload and desired outcomes and people will be able to advise better. If you’ve already done that then great.
2 Likes
I think we got to the right answer here for @Fastline, just more clarification on what happens with VDEVs. All my pools are single VDEV and I misunderstood how additional non-special VDEVs worked in a pool. Either way 2xRAIDZ2 pools is likely the best in this use case. Thanks for the clarification and explanation.
1 Like
I believe that is the case.
Yes, less resilver times as it has more disk which can increase the speed. More the VDEVs, more the performance, more IOPS.
That can be done i guess but hey, the reason why we migrated from multiple disks from the system to have one big pool and keep everything there in different directories unless you want to have two different pools for different use case scenario.
Yes, yes. Thank you for stepping in for further clarification. I needed this!
Yes, Thank you to everyone for the participation in this thread!