[Bug 275308] potential ZFS data corruption

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=275308

Olivier Certner <olivier.freebsd@free.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |olivier.freebsd@free.fr

--- Comment #6 from Olivier Certner <olivier.freebsd@free.fr> ---
From what I've read:

(In reply to Ed Maste from comment #1)
Given the patch itself, it's highly unclear it possibly can.  But it's true
that, in practice and so far, some people experiencing the issue (or one of
them, if there are several problems) have reported that effectively setting
'vfs.zfs.dmu_offset_next_sync' to 0 makes the problem impossible to reproduce
for them.

(In reply to Mark Johnston from comment #2)
It seems that setting 'vfs.zfs.dmu_offset_next_sync' to 0 actually *improves*
performance, at the expense of not reporting holes correctly for dirty files,
so potentially causing more processing for readers (and more storage
consumption in the case of, e.g., a concurrent 'cp').

-- 
You are receiving this mail because:
You are the assignee for the bug.