failed to read SMART Attribute Data

[Chris Moore]

I have 2 vdevs with 6 6TB drives in each. You're saying I could replace one drive in each vdev simultaneously.

The redundancy number of drives can be replaced. That means you could replace two drives in each vdev, but it would be a risk. If you had a drive failure, you would loose the pool, so I don't suggest it. The resiler time is the problem. You would likely have more than three days of resilver time per replace cycle and that time wold be due to the SMR feature.

With CMR drives, I am usually able to resilver in less than 24 hours with 12 TB drives.

My suggestion, since you have the drive bays for it, is to build a whole new pool of CMR drives and copy all your data over to the new pool. It should go smooth and fast. As fast as possible anyhow.

 

[kenster]

So yeah, if you can't attach additional disks, at least you can replace two at a time fairly safely.

yeah, so I actually can, with all the spare drive bays I have.

Is there an easy way to duplicate a pool exactly so I don't need to reconfigure all my smb shares and user permissions etc?

 

[kenster]

The redundancy number of drives can be replaced. That means you could replace two drives in each vdev, but it would be a risk. If you had a drive failure, you would loose the pool, so I don't suggest it. The resiler time is the problem. You would likely have more than three days of resilver time per replace cycle and that time wold be due to the SMR feature.

With CMR drives, I am usually able to resilver in less than 24 hours with 12 TB drives.

My suggestion, since you have the drive bays for it, is to build a whole new pool of CMR drives and copy all your data over to the new pool. It should go smooth and fast. As fast as possible anyhow.

thanks for all your helpful advice. this is likely what I'll do, and I guess I'll use WD Red Pro drives instead of WD Red this time. I did end up resilvering one drive before this dialog all finished, and it took about 18 hours.

 

[Ericloewe]

Is there an easy way to duplicate a pool exactly so I don't need to reconfigure all my smb shares and user permissions etc?

ZFS replication is easy-ish, but it's not going to be faster than resilvering. Just plug in all the new disks and replace them in one go. Zero juggling of pools is needed.
I'm not sure what is the correct way of doing this with the GUI, but if memory serves me right, in the CLI you just needed to issue them one at a time. The process will restart with every one you issue, but that's only a few minutes wasted.

 

[Chris Moore]

ZFS replication is easy-ish, but it's not going to be faster than resilvering. Just plug in all the new disks and replace them in one go. Zero juggling of pools is needed.
I'm not sure what is the correct way of doing this with the GUI, but if memory serves me right, in the CLI you just needed to issue them one at a time. The process will restart with every one you issue, but that's only a few minutes wasted.

In a RAID-z2 ?? I would never have thought to try replacing the entire pool in one go with a resilver. I have always established a new pool and used zfs send and receive. Is that resilver method something you have done?

 

[Ericloewe]

I haven't done it myself, but the crucial part, resilvering mulitple disks, is uncontroversial. My thinking is that send/recv is going to basically be doing the same as the resilver, since the objective is to copy everything over.
I think it's mostly a matter of people rarely having the means to attach n replacement disks in one go.

 

[kenster]

for what it's worth. I replaced the one complaining SMR drive, it took 18 hours to resilver. then I realized that a second drive in the same vdev was also complaining. I replaced it about 30 minutes ago, and it's already 25% complete resilvering.

but definitely looks like we need to move data asap. I have 12 drives on order now.

 

[kenster]

Just plug in all the new disks and replace them in one go. Zero juggling of pools is needed.

you're suggesting to replace all 6 drives of one vdev at once correct?

 

[Ericloewe]

Or the whole pool, if you can. A traditional resilver will take just a long regardless of how many disks you have, outside of weird edge cases. The recent sequential resilver (not to be confused with sequential scrub/resilver from a few years back) does allow for a faster rebuild by naïvely copying the whole disk over, but it needs to be followed by a scrub anyway. At that point, it's better to do a traditional resilver once and immediately fix whatever needs fixing, instead of redoing a ton of work multiple times.

 

[kenster]

Or the whole pool, if you can. A traditional resilver will take just a long regardless of how many disks you have, outside of weird edge cases. The recent sequential resilver (not to be confused with sequential scrub/resilver from a few years back) does allow for a faster rebuild by naïvely copying the whole disk over, but it needs to be followed by a scrub anyway. At that point, it's better to do a traditional resilver once and immediately fix whatever needs fixing, instead of redoing a ton of work multiple times.

right, using ZFS replication. I thouht you were talking about replacing the disks in the pool in place and resilvering.

 

[Ericloewe]

I thouht you were talking about replacing the disks in the pool in place and resilvering.

That is what I'm saying. You can just replace all the disks at once.

 

[kenster]

That is what I'm saying. You can just replace all the disks at once.

I think I'm with you now

replace all the drives in the existing pool and let it resilver as a blank pool. move the existing pool drives into 12 other slots and copy the data back to the original pool?