Date: Thu, 25 Sep 2014 16:40:56 -0400 From: Paul Kraus <paul@kraus-haus.org> To: freebsd-questions@freebsd.org Subject: Re: zfs; adding vdev with same size and raid level but different layout Message-ID: <54247DD8.2060107@brandywine.kraus-haus.org> In-Reply-To: <E4AEEF9C-D8AF-4124-BEB1-B3C3C316D140@gmail.com> References: <E4AEEF9C-D8AF-4124-BEB1-B3C3C316D140@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 9/25/14 16:09, Stefan Johansson wrote: > I’m running FreeBSD 9.3-RELEASE with a zpool in raidz1 with 5 2TB disks. > Now I’m planning to expand the pool with a vdev with the same effective size (8TB). > Would it be a bad idea to add 3 4TB disks in raidz1 to the pool instead of another 5 2TB disks? > What could be the disadvantage with such a setup? The major disadvantage would be performance difference between the two vdevs, but since most RAIDz1 performance limits are due to the speed of one drive in the vdev, that should not be a major difference. Remember to have a large enough hot spare (or even a cold spare on the shelf) so that _when_ you lose a device you are not waiting a long time to find out or replace it. With large zpools you are more vulnerable to multiple device failures due to the longer resilver times. Unlike hardware RAID, ZFS resilver times are NOT based on raw sequential performance of the devices, but are limited by host CPU and random I/O limits of the devices since a resilver is more like a walk of the zpool in time replaying transactions. You can even mix up vdev types within a zpool (RAIDz1, 2-way mirror, RAIDz2, etc.). The biggest disadvantage that has been discussed over on the ZFS list is that you really can't predict the performance of such a "hybrid" zpool at all since you never _know_ which vdevs will be handling the write/read. -- -- Paul Kraus paul@kraus-haus.org Co-Chair Albacon 2014.5 http://www.albacon.org/2014/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?54247DD8.2060107>