Sanity check of filesystem design

I'm contemplating a major overhaul of the filesystems on my Ultra 20 and would like a sanity check before I start reconfiguring the drives.

I've printed and read the relevant manuals and several how to's but it's a lot to read, so I'm concerned I might have missed something. I've also never played with SVM before. I'm already using ZFS and it seems pretty straight forward.

System is running Solaris 10 5/08 and has 2 x 500 GB Maxtor drives. The geometries are NOT the same. One is 60800 cylinders x 16065 sectors the other 60563 cylinders x 16128 sectors.

What I have in mind is as follows:

Disk 0 slices:

4 boot environment slices for live upgrade

1 swap

1 ZFS pool mirror for home

1 ZFS pool slice (not mirrored) for data


Disk 1 slices:

1 boot slice mirror

1 ZFS pool mirror for home

1 ZFS pool slice (not mirrored) for data

The idea being that I'll break the mirror whenever I change boot environments, do an upgrade, etc and rebuild it.

Home will be a ZFS mirror and data will be on an unreplicated filesystem created by ZFS from the slices on the 2 drives. I realize I'll lose a little space in the mirror set because of the geometry mismatch, but it appears to me from the manual that this will work, though some of the how to's emphasize having identical drive geometries.

The home filesystem will be sized to match the size of the home on my laptop so I can easily sync the two. I'm not concerned about having automatic failover. What I'm looking for is being able to restore operation quickly in the event of a device failure. The only reason for mirroring root is to simplify keeping the 2nd BE in sync w/ the 1st.

Does this make sense to those who have worked w/ SVM & ZFS? Is there something radically wrong w/ the scheme? Obviously ZFS root would be better, but it's my primary work system, so as good as OpenSolaris is, I'm not quite ready to make that jump.

Thanks,
Reg
.

Relevant Pages

  • Re: Sanity check of filesystem design
    ... I'm already using ZFS and it seems pretty straight forward. ... boot environment slices for live upgrade ... ZFS pool mirror for home ... Home will be a ZFS mirror and data will be on an unreplicated filesystem created by ZFS from the slices on the 2 drives. ...
    (comp.unix.solaris)
  • Re: Sanity check of filesystem design
    ... I'm already using ZFS and it seems ... boot environment slices for live upgrade ... ZFS pool mirror for home ... mirroring root is to simplify keeping the 2nd BE in sync w/ the 1st. ...
    (comp.unix.solaris)
  • Re: UFS to ZFS
    ... it over to ZFS? ... to do is ZFS the swap (slice 0 on both drives) and /export/home ... Do you intend to mirror them on ZFS as well? ...
    (comp.unix.solaris)
  • Re: Would ZFS and gmirror work well together in a two-node failover cluster?
    ... I want ZFS to ... manage the drives as a JDBOD in a RAIDZ2 configuration. ... real time mirror is with gmirror and ggated, ... gmirror would interact with ZFS. ...
    (freebsd-current)
  • Re: zfs - move pool to new device?
    ... mirror vdevs and expected reads to only occur from two drives at a ... time (single half of a mirror). ... Since ZFS is smart enough to know ... yes but all your drives are likely identical and on the same kind of ...
    (comp.unix.solaris)