Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 10 Aug 2015 09:20:35 -0400
From:      PK1048 <paul@pk1048.com>
To:        FreeBSD Filesystems <freebsd-fs@freebsd.org>
Subject:   Re: Urgently need some help solving lost space due to snapshots during receive op
Message-ID:  <43B00232-46B7-4B96-B80C-D0111C6D8DA2@pk1048.com>
In-Reply-To: <55C863DE.3000200@digiware.nl>
References:  <55C82BCD.2050404@herveybayaustralia.com.au> <55C863DE.3000200@digiware.nl>

next in thread | previous in thread | raw e-mail | index | archive | help
On Aug 10, 2015, at 4:42, Willem Jan Withagen <wjw@digiware.nl> wrote:

> Last number I remember as sensible maximum fill level is around 70%.
> If you go over it, your system is going to be busy with ZFS =
bookkeeping
> instead of data storage.

Fortunately or not, that number is _very_ dependent on your specific =
data and how it is loaded. If it is static data, loaded once and not =
modified, you can run a much higher % Used, for lots of changes on an =
ongoing basis you need much more free space. The issue is that since ZFS =
is Copy on Write, every write (even a modification to an existing file =
or disk block) needs to find space. ZFS tries to reduce fragmentation. =
As the zpool gets more and more full it takes longer and longer to find =
the best possible location to put the new write. It also becomes harder =
and harder to find contiguous blocks to put the new write into. The =
existence of snapshots makes this problem even worse, as previously =
written data is not cleared until the snapshot is destroyed.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43B00232-46B7-4B96-B80C-D0111C6D8DA2>