Hot spares

32 views
Skip to first unread message

Anders Wallén

unread,
Apr 14, 2013, 2:55:54 PM4/14/13
to zfs-...@googlegroups.com
Greetings, everybody!

I am new to the ZFS world and have just set up my first array. Once I got the commands figured out, it was a snap and went without problems.

Something I could not work out on my own, which is why I am making this post, is how to make a hot spare for the pool. The cabinet I use has eight bays, and all eight discs are ZFS-formatted, but the last one is not yet attached to the pool.

Is there any way to make the server automatically make use of the 8th disc when another one fails?


Best regards,

Anders

Jason

unread,
Apr 14, 2013, 3:17:52 PM4/14/13
to zfs-...@googlegroups.com
You can add it as a spare, BUT it will not auto swap. I have 2 such pools and at least the spare is present and ready when needed with a simple command. So far, no issues, manually works well. Future updates will resolve this.....

Jason
Sent from my iPhone
> --
>
> ---
> You received this message because you are subscribed to the Google Groups "zfs-macos" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to zfs-macos+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

Lucien Pullen

unread,
Apr 14, 2013, 5:05:31 PM4/14/13
to zfs-...@googlegroups.com
Also sprach Jason at 4/14/13 2:17 PM:
> You can add it as a spare, BUT it will not auto swap. I have 2 such pools and at least the spare is present and ready when needed with a simple command. So far, no issues, manually works well. Future updates will resolve this.....

Could you add this to the issue tracker and include your workaround?
Should the documentation be updated to reflect this, or just keep it on
the issue tracker? The documentation currently says the hot spare will
be automatically replaced. Either way, I think adding the workaround in
EXAMPLES would be a good idea.

Daniel Becker

unread,
Apr 14, 2013, 5:08:35 PM4/14/13
to zfs-...@googlegroups.com
AFAIK auto-replacement is handled by an external daemon on Solaris, not by ZFS itself. I don't think hot spares currently work correctly on any of the third-party implementations.

Jason

unread,
Apr 14, 2013, 5:18:42 PM4/14/13
to zfs-...@googlegroups.com
Interesting. I've had the issue come up twice. Neither time did the spare auto swap and I had assumed from much debate that this was the current approach in OS X. Happy if I'm wrong.

Jason
Sent from my iPhone

Jason

unread,
Apr 14, 2013, 5:20:02 PM4/14/13
to zfs-...@googlegroups.com
Right. That confirms what you and I have chatted about in the past. And my experience. ;)

Jason
Sent from my iPhone

Chris Ridd

unread,
Apr 15, 2013, 2:54:19 PM4/15/13
to zfs-...@googlegroups.com

On 14 Apr 2013, at 22:20, Jason <jason...@belecmartin.com> wrote:

> Right. That confirms what you and I have chatted about in the past. And my experience. ;)

I read a blog post this morning about ZFS that strongly advised against using hot spares on (I'm assuming) Nexenta. IOW an OS likely to handle them better than OS X.

http://nex7.blogspot.co.uk/2013/03/readme1st.html

I couldn't see anything explaining that advice.

Chris

Richard Elling

unread,
Apr 15, 2013, 3:07:42 PM4/15/13
to zfs-...@googlegroups.com
On Apr 15, 2013, at 11:54 AM, Chris Ridd <chris...@gmail.com> wrote:

On 14 Apr 2013, at 22:20, Jason <jason...@belecmartin.com> wrote:

Right. That confirms what you and I have chatted about in the past. And my experience.  ;)

I read a blog post this morning about ZFS that strongly advised against using hot spares on (I'm assuming) Nexenta. IOW an OS likely to handle them better than OS X.

Yes, that is a blog from a Nexenta employee.
No, I do not always agree with him :-)


http://nex7.blogspot.co.uk/2013/03/readme1st.html

I couldn't see anything explaining that advice.

This has nothing to do with ZFS. It has everything to do with the way Solaris manages
faults, with its default settings. AFAIK, Nexenta does not change these defaults, so they
are similarly affected. Clearly, OSX is not Solaris and therefore must have a different
mechanism for declaring a retirement.

That said, in my ZFS tutorials, I go through some analysis that explains how you can
determine whether hot spares are worthwhile. In general, more parity is better than
less parity + spares. Intuitively, this makes sense if you think of a hot spare as an 
unsilvered parity disk... if you keep it silvered, you don't have to resilver it after a 
failure. This intuition works for smallish deployments, but you'll need to do some math
for configs with more than a few dozen disks.

The other angle is that for sites providing 7x24 support, hot spares don't offer much value.
Hot spares are great for sites where the staff goes home over the weekend.
 -- richard


Jason

unread,
Apr 15, 2013, 3:08:35 PM4/15/13
to zfs-...@googlegroups.com
I have several systems (OS X) with and without and no issues other than you manually having to swap on those that do. The big plus to having them there is that it's quick to fix and I can physically collect and replace the bad drive later. If one doesn't have one in the rack and ready, it just takes longer. ;)

Jason
Sent from my iPhone

Jason

unread,
Apr 15, 2013, 3:09:47 PM4/15/13
to zfs-...@googlegroups.com
Yes. Good explanation. As always. 


Jason
Sent from my iPhone
Reply all
Reply to author
Forward
0 new messages