Jacob Yocom-Piatt wrote:
lo all,

have a freebsd 7.0-beta4 machine attached to an external disk enclosure and would like feedback on the following setup: have RAID-Z on 4 disks, ZFS volume that takes up entire RAID-Z, use ZVOL from volume for encryption via geli, use .eli (decrypted) device to make another ZFS pool. the idea being "no time/resources wasted doing fscks plus encryption sans hardware RAID".

Unless I'm misunderstanding your objective, geli'ing the disks and creating a pool ontop of the encrypted disks (zpool create secure raidz da0.eli da1.eli da2.eli da3.eli) would also work, and would be far easier to maintain. Your data would still be encrypted and you would still need to provide the passphrase to make the pool accessible.

Best regards,


translated to commands this reads:

# zpool create p_a raidz /dev/mfid1 /dev/mfid2 /dev/mfid3 /dev/mfid4
# zpool list
NAME                    SIZE    USED   AVAIL    CAP  HEALTH     ALTROOT
p_a                    2.72T   4.02G   2.71T     0%  ONLINE     -
# zfs create -V 2048g p_a/vol
# geli init -K /root/p_a.key -s 4096 -l 256 /dev/zvol/p_a/vol
# geli attach -k /root/p_a.key /dev/zvol/p_a/vol
# zpool create a /dev/zvol/p_a/vol.eli

i got a reboot while scp-ing some files to /a (only got ~3 GB in) from another machine with the above setup. am currently waiting far too long for a rm -R <directory> to complete under /a. will test if any of this behavior is repeatable.

i welcome opinions or suggestions on the stability of such a setup (ZFS-geli-ZFS) and if this is not stable, as the reboot i just experienced would indicate, suggestions on alternative configurations that allow use of geli and minimize or eliminate fsck time. i do have a preference for no hardware RAID since it ties us to a particular card. will furnish a proper bug report if the reboots are repeatable in the aforementioned scenario.

NOTE: please CC me since i am not yet subscribed to this list


freebsd-questions@freebsd.org mailing list
To unsubscribe, send any mail to "[EMAIL PROTECTED]"

Reply via email to