Quoting Tim Spriggs <[EMAIL PROTECTED]> on Fri, Mar 28 12:08:
>
> raidz root will work in Nexenta when it works in vanilla Open Solaris. 
[ snip ]
> The zfs-discuss archives might have more information on this.

Good to know, and thanks for the pointer, I will definitely check the
list out.

> I would suggest that you keep your entire root filesystem on the 
> mirror. apt-clone relies on this assumption in order to rollback 
> packaging decisions and does not snapshot/clone other filesystems.

This raises a couple more questions about how pools work (I may be
wondering into ZFS only territory here).  Say I create a 100GB 3-way
mirror, and that becomes syspool.  The other 900GB disk partitions would
go into a raidz.  Would that be a new pool with a name of my choice?  Is
there an easy upgrade path when Open Solaris supports raidz boot besides
wipe and re-install?  From what I have read I could drop one disk out of
the raidz, re-add the entire disk into the raidz, and repeat for the
other 2 disks?  Would that work?

Also, would I need to make the 3x100GB partitions be a straight mirror,
or could I make it a pool and tell ZFS to keep at least 2 copies of the
data, that way it would survive 1 disk failing, but is still a mirror
(of sorts).  Can GRUB boot from that kind of "mirror"?  Is it fully
robust against a single disk failure?  From the zfs blogs it seems as
if both copies could end up on the same disk...

Quoting Erast Benson <[EMAIL PROTECTED]> on Fri, Mar 28 12:16:
>
> I calculated this number recently - your syspool need to have ~
> (DownloadSize x 6)  of free space..
> 
> For typical ON upgrade, it could be 600MB...900MB. To be on a safe side,
> always check that 'zfs list syspool' command shows >= 1GB of AVAIL

Excellent to know, thanks for the info.  

I seem to be missing an acronym, what is ON?

Thanks again!

-- 
Real Users hate Real Programmers.
_______________________________________________
gnusol-users mailing list
[email protected]
http://lists.sonic.net/mailman/listinfo/gnusol-users

Reply via email to