Possible data corruption of incremental backups "ZFS send fails to transmit some holes" OpenZFS Bug

Status
Not open for further replies.
Joined
Mar 12, 2016
Messages
3
Hello,

from what i found out just today, there is a bug that was introduced a while back and is to my knowledge still affecting all (relatively) new FreeNAS versions, as it was only fixed in the last week(s). 9.3.x should, i guess, at least be affected.

This bug would be concerning a feature named "hole_birth", which is part of newer zfs pool versions. As far as I was able to ascertain, deactivating this feature after data has been placed and or zfs send/resv was used, could further worsen matters.

From what i could gather, it pertains to incremental zfs sends, where, instead of a set of zeros, data from older versions of files is included, thus creating data corruption in certain cases for the incremental backups.

Someone from the #freebsd channel said, that the issue was fixed in 10.3-RC2, although unconfirmed.

It should be part of the next zfs versions released by illumos and zfs on linux.


What i hope someone can answer me is this: When and exactly how is the team behind FreeNAS planning to resolve this - from my perspective rather problematic - issue?

Thank you in advance for your help.

Best regards.
 
D

dlavigne

Guest
You won't get that answer on the forums. Please create a bug report at bugs.freenas.org that contains this information and post the bug number here.
 
Joined
Mar 12, 2016
Messages
3
I was assuming, that the bug was known and the work underway. Therefore I was solely interested in knowing what the plans to include the bug-fix were.
 
D

dlavigne

Guest
Again, the devs don't hang out in the forums and they are the ones who would know... You could also try the IRC channel to see if one is around, or create a bug report to track its progress (or to be told whether or not it's already fixed or if it even affects the FreeBSD version).
 
Status
Not open for further replies.
Top