Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 4 Dec 2014 14:52:51 -0800
From:      Craig Yoshioka <craigyk@nanoimagingservices.com>
To:        freebsd-fs@freebsd.org
Subject:   remove or make reserved ZFS space configurable
Message-ID:  <1D872444-CF75-48FF-BFDE-51885A3BBF9B@nimgs.com>

next in thread | raw e-mail | index | archive | help
I saw an earlier thread about an update that added (invisible) reserved =
space for ZFS Pools.

I use ZFS as the filesystem for my backup disks.  This allows me to =
occasionally rotate through old backup drives and scrub them to detect =
damage without having to implement such detection in my backup system.  =
I can then re-backup bad files, or reconstruct a backup drive, if =
necessary.   I configured my backup software to leave ~100GB free (3TB  =
drives) and I was not amused when I recently mounted a set of backup =
drives, and puzzled over why they were now "full=E2=80=9D and if I had =
messed up and lost data.  And even with the reserved space, while =
troubleshooting I tried a rm and it failed (so what's the point of the =
reserved space?).

I don't care about the performance degradation on these drives... these =
drives are 99% single-write.  I agree, that having reserved space as a =
default is probably good, but why isn't this implemented as a =
configurable option (like a default zpool/zfs quota property)?  Instead =
it looks like I have to wait for an update to make it to release and =
then set some kernel option?  To prevent accidental filling of a pool, I =
usually create a zfs filesystem on each pool with about ~10% reserved =
space, is this solution not workable?
=20
Thanks,
-Craig







Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1D872444-CF75-48FF-BFDE-51885A3BBF9B>