[Gluster-users] replacing brick procedure

[Lindsay Mathieson]

On 21 April 2016 at 15:01, Bishoy Mikhael <b.s.mikhael at gmail.com> wrote:

> you don’t need to create a directory and set the extended attributes
> manually anymore if you are using Gluster 3.7.x.
>


Thats really good to know, thanks.



> But the question is, why do you need the replace brick command if you are
> using ZFS?!
>


Heh!

Last night I had a disk in a Raid10 array fail. Then the 2nd disk in the
*same* mirror started throwing errors! a bad moment. I had no chance of
getting replacment disks until the next day so I detached the disk from the
2nd mirror and replaced the first mirror with it, converting the RAID10 to
a RAID0, not ideal but the best option I had. Fortunately the re-sliver
finished before the 2nd disk packed it in. Tonight I'll be adding two new
disks to convert it back to RAID10.

Dog bless ZFS and its checksums, would have been screwed without them.

Both disks in a mirror failing is rare, but it can happen :) so I thought
I'd brush up on my replace-brick procedure.

nb: The reduction in IOPS from RAID10 to RAID0 is quite noticeable.


Also, despite what you may have read on the internet, never use desktop
drives in a server setup :)
-- 
Lindsay
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160421/059d76af/attachment.html>