Project

General

Profile

Bug #11840

Remove of a special vdev with different ashift than the pool vdevs results in an OmniOS panic/pool corrupt

Added by Guenther Alka 6 months ago. Updated 6 months ago.

Status:
New
Priority:
Urgent
Assignee:
-
Category:
-
Start date:
Due date:
% Done:

0%

Estimated time:
Difficulty:
Medium
Tags:

Description

I tried to remove a special vdev build from an Optane partition.
On a zpool remove the OS crashed immediately and after reboot during pool mount

Crash dump, see napp-it.org/doc/downloads/omniosce.zip

The problem seems to happen only when you add a special vdev with a different ashift than the ashift of pool vdevs.
Without forcing ashift, the Optane partitions as special vdev defaults to ashift=9

When I force ashift=12 for the special vdev everything is ok.
The solution could be that a special vdev remove should not proceed with different ashift.
This is the behaviour on a regular vdev remove

My test environment
ESXi 6.7u3 with OmniOS bloody (october), HBA with a pool from a basic HGST HE8 and Optane 900 pass-through


Related issues

Has duplicate illumos gate - Bug #11851: ZFS special vdev ashift mismatch causes panic on removalClosed

Actions

History

#1

Updated by Gergő Mihály Doma 6 months ago

  • Has duplicate Bug #11851: ZFS special vdev ashift mismatch causes panic on removal added
#2

Updated by Matthew Ahrens 6 months ago

Also available in: Atom PDF