Time difference for resilvering two drives in different vDevs?

ITOperative

Dabbler
Joined
Feb 11, 2023
Messages
20
Good afternoon!

I just got some drives in to replace da10 (in vDev2) as it has been failing Long SMART tests, but today I noticed I just got a Long SMART failure on da1 (vDev1) too.

I wanted to confirm three things before proceeding and figured I'd ask the community:
  1. Should I outright replace a drive for long SMART failures or is there any command(s) I should run first to confirm how bad it actually is?
  2. If I replace two drives, each in two different Raidz2 vDevs (one drive per vdev), how does this impact silvering time? Does it double the time taken, or does ZFS take that into account and reduce the time taken?
  3. Is it a bad idea to handle both replacements at the same time?
Thanks so much!
 

Alecmascot

Guru
Joined
Mar 18, 2014
Messages
1,177
Last edited:

ITOperative

Dabbler
Joined
Feb 11, 2023
Messages
20
HGST SAS drives can fail the long test in the second segment if the host interrupts.
This is a firmware issue...
Does that mean there's nothing wrong with the drive, or do you mean the firmware on the drive's controller?

If the former, would that more likely be motherboard firmware or HBA firmware?

Thanks for the quick reply!

EDIT:
Oh, and I should note specifically da10 has had like 3 out of 3 long SMART failures so far, and da1 has had 2 out of 2 so far.
The other 15 drives seem to be fine so far.
 

Alecmascot

Guru
Joined
Mar 18, 2014
Messages
1,177
I added a link.
Drive firmware.
 

ITOperative

Dabbler
Joined
Feb 11, 2023
Messages
20
I added a link.
Drive firmware.
It would seem both da1 and da10 are already running revision A3A0, and they also seem to both be failing on segment 7 instead.
My other drives are almost all running A3A0 as well, with a couple running A320.

Interestingly, only two are running C250, and while unrelated to da6 and da10, the two C250 drives actually won't SMART long test, just short test, so I may look into their firmware.
 
Top