Re: ZFS RAIDZ2 and wd uncorrectable data error - why does ZFS not notice the hardware error?

[mlelstv%serpens.de@localhost (Michael van Elst)]

mp%petermann-it.de@localhost (Matthias Petermann) writes:

>wedges at boot time unnecessarily endangers the RAID in the event of a=20
>disk change. Therefore the question: is there a better possibility=20
>besides using the wedges? I remember that I had also tried the variant=20
>with the label NAME=3Dzfs2 when I created it (as it works with newfs, for=
>=20
>example), but it didn't work. Ok - as a workaround I could have prepared =

>the disk on another system - for the next time I know that now.


One solution is to use devpubd to create symlinks in /dev/wedges, e.g.

lrwxr-xr-x  1 root  wheel  8 Jul 17 22:03 /dev/wedges/0d2c1666-075f-4a70-bd04-c3f2913fbc80@ -> /dev/dk7
lrwxr-xr-x  1 root  wheel  8 Jul 17 22:03 /dev/wedges/41993b4b-180c-4df3-bd6b-25b5cbc036cb@ -> /dev/dk6

The zpool can then be created using these links.

For this to work, devpubd needs to run before zfs by changing the
start script from 

# REQUIRE: DAEMON
# BEFORE:  LOGIN

to e.g.

# REQUIRE: root
# BEFORE:  zfs ccd cgd lvm raidframe

The devpubd hooks also used commands from /usr, then (with netbsd-9
or older -current) /usr must not be a separate partition. I have
updated the hooks in -current to avoid /usr, but for custom hooks
that still might be an issue.

N.B. recent -current is broken, zfs cannot access such symlinks.