Replace failed hdd issue

[iliak]

i have a raidz1 with 8 ssds
one had some issues and was marked as unavailable.
i have set it to offline, swapped and selected replace.
after some time, Pool status still have errors and it stays in degraded,
what could be the reason ? should i try to replace it again?

 

[Heracles]

Hi illiak,

RaidZ1 is to be avoided and you may be another example of it...

When a drive fails in a RaidZ1, the pool ends up with no protection at all. As such, if there are other errors anywhere in the pool or some happen during the re-silvering process, there is no way for ZFS to recover from these. So one possibility is that : your pool was not perfectly clean at 100% so the re-silvering process was unable to recover. With 8 drives, you should use RaidZ2 to protect yourself against such a case.

With only 7 errors, it may very well be your situation.

You can try to replace the drive again in case you ended up with an already defective replacement, but know that it may be too late. If you do not have a backup of your data, do one ASAP, even before trying the second replace.

Once the backup is done and second replacement is tried, you should also re-design your setup and stop using RaidZ1.

Good luck,

 

[iliak]

can i force replace without corruption test? to ignore the errors?
( i know it is not safe) but i need to do something until the weekend (ill do the change when)

 

[Heracles]

Not sure what you mean by replacing without corruption test... ZFS has built-in integrity protection and that integrity is everything why we use ZFS. To use ZFS without integrity is like asking to drive a car without an engine...

So No, you can not disable ZFS native integrity check. The one thing to do right now is to backup your data for you to save whatever can be saved (probably almost everything as of now but don't waste time).

 

[iliak]

first of all all the data is recoverable, we have multiple copie, this server is used only to improve performance and remove lots of load from our main server..

one more question
can i do a snapshot from RaidZ1 and load it back to RaidZ2?

 

[Heracles]

Good for you that you have a backup.

You sure can do that : Pool-A is RaidZ1, you take a snapshot, configure replication to Pool-B which is RaidZ2 and once received by Pool-B, that snapshot will be hosted in a RaidZ2 with all the benefits.

 

[iliak]

i cant do it yet, we have only 1 freenas server and all the hdds are used for this pool, i figure it will be easier just to remake it, it is a simple setup,
the setup is simple ill just remake it.

 

[Ericloewe]

No, you can not disable ZFS native integrity check

Sure you can, but there are very few good reasons to do so, none of them in production. checksum=disabled or whatever.

 

[Heracles]

Hi Eric,

Sure you can, but there are very few good reasons to do so, none of them in production. checksum=disabled or whatever.

Ok... I will take note on that one but that really looks like something people are best not to know about