Redundancy mismatch on new pool

[niah]

Hi all,

Destroyed an old RAIDZ1 pool and moved to a new RAIDZ2 pool, but after creating the new one I have a Redundancy Mismatch warning on my mirrored metadata VDEV. I've tried restarting, recreating the pool, creating the pool with just a data VDEV and adding the metatdata after creation, removing and re-adding disks after creation, but didn't have any luck.

My layout is:
3x RAIDZ2, 6 wide Data
1x Mirror, 2 wide Metadata

I found this thread and it looks like this (or something similar) might have been a visual bug at some point. Is this related?

Redundancy mismatch after applied the latest update

Applied the update yesterday and there are two things pop up for one of my pools. One is a failed SMART test at lifetime 11(aborted due to restart) and with some search it seems it will eventually be flushed with newer records as long as I use the drives for enough time, they are brand new...

www.truenas.com

 

[HoneyBadger]

Hey @niah

In this case, the message is correct to indicate a redundancy mismatch; your data vdevs can survive the loss of two disks each, but the metadata is only a two-way mirror and thus can only afford the loss of a single drive.

You can either ignore the error (as it doesn't impede functionality) with the understanding that losing both meta drives means an unavailable pool, or add a third meta drive to the same mirror volume with the EXTEND option in the UI to give it two-disk failure tolerance.

 

[niah]

Ah thanks. I was just misunderstanding what the warning was for.

 

[Michael Schefczyk]

The warning also made me think, but I am not certain if I did understand everything correctly ... My layout is:

Data VDEVs 1 x RAIDZ2 | 4 wide | 18.19 TiB
Metadata VDEVs 1 x MIRROR | 2 wide | 1.82 TiB
Log VDEVs 1 x MIRROR | 2 wide | 93.16 GiB
Cache VDEVs 1 x 1.82 TiB

I am happy with this layout. It would also be no problem to add one or two more metadata SSDs of the same size.

Redundancy mismatch does not imply that the layout has to match, but that Data and Metadata need to withstand the same number of disks failing, correct?

I will not be possible to add 2 SSDs and migrate the metadata VDEVs from MIRROR to RAIDZ2, also correct?

Adding one SSD to the metadata VDEV while keeping it as mirror will make the warning go away, right?

Thaks a lot!

 

[HoneyBadger]

Redundancy mismatch does not imply that the layout has to match, but that Data and Metadata need to withstand the same number of disks failing, correct?

Correct. In your case, DATA can lose two disks (RAIDZ2) and METADATA only one.

I will not be possible to add 2 SSDs and migrate the metadata VDEVs from MIRROR to RAIDZ2, also correct?

Also correct. Metadata can only be STRIPE or MIRROR types, it does not support RAIDZ of any width.

Adding one SSD to the metadata VDEV while keeping it as mirror will make the warning go away, right?

Correct a third time. Use the Storage -> Manage Devices pane of the UI. Select your existing metadata vdev, then use the "EXTEND" option and select your third SSD. It will resilver the data from the other two and resolve the warning.

 

[Michael Schefczyk]

Thanks a lot!!