[zfs-discuss] Replacing disks best practices

Gregor Kopka (@zfs-discuss) zfs-discuss at kopka.net
Mon Nov 26 05:11:02 EST 2018

Am 26.11.2018 um 09:07 schrieb Gionatan Danti via zfs-discuss:
> On 23/11/2018 15:02, Sietse van Zanen wrote:
>> My preferred plan is to have somebody buy a new drive, so I don't
>> have to pay for it and then have somebody else replace it, so I don't
>> have to do anything. This plan doesn't always work though. :)
>> Your question is really a personal preference. autoreplace is nice,
>> but it may also automatically replace something that should not be
>> replaced or replace it with something that should not be used as a
>> replacement.
>> If you take care of wich and how drives are replaced than use
>> autoreplace. If you want more control over the process to decide what
>> you replace with what exactly, than do it manually.
> Hi, thanks for your reply.
> Anyone can comment regarding the "offline before replace" thing? Do
> you offline your disks before replacing? Or do you simply "pull the
> plug" on the bad disk, without offlining it?
In case the drive has only some bad sectors as the spare ones are
exhausted and isn't otherwise slow I leave it in for the replacement as
it stil functions as redundancy (so it keeps chance of pool death
through additional failures a bit lower). In case it's having trouble to
read in general (thus slowing access to the pool) I offline it to speed
up the replacement.


More information about the zfs-discuss mailing list