Project

General

Profile

Bug #766

RAIDz spare drive issue

Added by Roy Sigurd Karlsbakk over 8 years ago.

Status:
New
Priority:
Normal
Assignee:
-
Category:
kernel
Start date:
2011-03-04
Due date:
% Done:

0%

Estimated time:
Difficulty:
Tags:

Description

Hi all

If having a RAIDz2 pool, tested with a single-VDEV pool with two spares, removing two drives from the pool, waiting for the pool to resilver and then unplugging a third of the original VDEV members, keeping the spares, all pools on the system will hang, possibly until a reboot. This was also tested on a mirrored pool (seven or so VDEVs and one spare) where I removed first one of the drives in VDEV 1, waited for resilver to complete, and removed the second side of the mirror for that VDEV. As the worked well, I guess this bug is limited to RAIDz use. I haven't yet tested this on RAIDz13. Details of how to reproduce this is found here http://karlsbakk.net/ZFS/ZFS%20Spare%20disk%20usage.pdf

I tested this on both Hitachi Deskstars and WD Green drives, on both an LSI 8601 and an older AOC-SAT2-MV8. On the former, a 24-drive box from Supermicro, I made sure to test all VDEVs locally and remote after the removal of the third drive. Neither worked, and reconnecting the drive didn't help much either. I ended up rebooting and destroying the pool.

I did some minimal testing also on S11Ex, but from the live session, S11Ex didn't trigger any resilver to spares, neither with the original v28 zpool version, nor on whatever version S11Ex uses (I did an upgrade and some initial testing before I destroyed that pool).

Also available in: Atom PDF