Very Computer

Solaris 9

I have metadevices that I cannot get rid of.

Is there a way to wipe out all SVM devices by say ... deleting all metadb
state repliaces in both the local state database and diskset state
database ?

What about deleting disksets in an unconventional way ?? At the moment, I
cannot even access the disksets, as I am getting "metaset: syddb280r: No
such device".

So I would really like a way to wipe out all SVM devices and disksets
WITHOUT having to reinstall Solaris.

1) Cannot take ownership or delete a metaset:

# metaset -s testset

Set name = testset, Set number = 2

Host Owner
syddb280r
syddbe220r

Drive Dbase

c3t0d5 Yes

c4t0d5 Yes

# metaset -s testset -t -f
metaset: syddb280r: testset: there are no existing databases

bash-2.05# metaset -s testset -t
metaset: syddb280r: testset: there are no existing databases

bash-2.05# metaset -s testset -d c3t0d5
metaset: syddb280r: testset: must be owner of the set for this command

2) The device ID is no longer up to date with SVM ... eventually
ending up with "metastat: No such device "

$ metastat -s shared3310 d50

shared3310/d50: Mirror
Submirror 0: shared3310/d51
State: Needs maintenance
Submirror 1: shared3310/d52
State: Needs maintenance
Pass: 1
Read option: roundrobin (default)
Write option: parallel (default)
Size: 204767040 blocks (97 GB)

shared3310/d51: Submirror of shared3310/d50
State: Needs maintenance
Invoke: metasync shared3310/d50
Size: 204767040 blocks (97 GB)
Stripe 0:
Device Start Block Dbase State Reloc Hot
Spare
c3t0d4s0 0 No Okay Yes

shared3310/d52: Submirror of shared3310/d50
State: Needs maintenance
Invoke: metasync shared3310/d50
Size: 204767040 blocks (97 GB)
Stripe 0:
Device Start Block Dbase State Reloc Hot
Spare
c4t0d4s0 0 No Okay Yes

Device Relocation Information:
Device Reloc Device ID

< ... snip ...>

I tried to detach the submirrors, but I then get the
following errors:

bash-2.05# metadetach -s shared3310 -f d50 d52
metadetach: syddb280r: shared3310/d50: resync in
progress

bash-2.05# metadetach -s shared3310 -f d50 d51
metadetach: syddb280r: shared3310/d50: resync in
progress

# metadb -i -s shared3310
flags first blk block count
a m luo 16 8192 /dev/dsk/c3t0d0s7
a luo 16 8192 /dev/dsk/c3t0d1s7
a luo 16 8192 /dev/dsk/c3t0d2s7
a luo 16 8192 /dev/dsk/c3t0d3s7
M 16 unknown /dev/dsk/c3t0d4s7
a luo 16 8192 /dev/dsk/c4t0d0s7
a luo 16 8192 /dev/dsk/c4t0d1s7
a luo 16 8192 /dev/dsk/c4t0d2s7
a luo 16 8192 /dev/dsk/c4t0d3s7
M 16 unknown /dev/dsk/c4t0d4s7

# metadb -s shared3310 -d c4t0d4s7

# metadb -s shared3310 -a c4t0d4s7

# metadb -i -s shared3310
flags first blk block count
a m luo 16 8192 /dev/dsk/c3t0d0s7
a luo 16 8192 /dev/dsk/c3t0d1s7
a luo 16 8192 /dev/dsk/c3t0d2s7
a luo 16 8192 /dev/dsk/c3t0d3s7
M 16 unknown /dev/dsk/c3t0d4s7
a luo 16 8192 /dev/dsk/c4t0d0s7
a luo 16 8192 /dev/dsk/c4t0d1s7
a luo 16 8192 /dev/dsk/c4t0d2s7
a luo 16 8192 /dev/dsk/c4t0d3s7
a u 16 8192 /dev/dsk/c4t0d4s7

# metastat -s shared3310 d50
metastat: syddb280r: No such device

# metastat -s shared3310 d30
metastat: syddb280r: No such device

# metastat -s shared3310
metastat: syddb280r: No such device

# metaset -s shared3310
metaset: syddb280r: No such device

Not one of your commands seesm to have tried adding the -f flag to your
arguments, i.e. force it whatever the error.

IF that fails, then we can investigate further.

IF you cant remove the metasets, then you can remove the "local set" and
reinitiate that, that should wipe any knwledge about the metasets from it.

/Johan A

Quote:

> IF that fails, then we can investigate further.

> IF you cant remove the metasets, then you can remove the "local set" and
> reinitiate that, that should wipe any knwledge about the metasets from it.

If I then recreate the disksets, I presume SVM will not read any existing
diskset state database replicas that previously existed on the disks.

Since disksets have their own metadbs, and they are not listed
in /etc/lvm/* , ... SVM must know of the location of the state replicas
for disksets from the local metadb

Having said that, you actually said:

== IF you cant remove the metasets, then you can remove the "local set" and
== reinitiate that, that should wipe any knwledge about the metasets from
it.

I actually originally read that you meant "remove the local state replicas".
Meaning, I presume you meant "metadb -d", without specifying any diskset
parameter, until all local state replicas are deleted, will effectively
wipe out any knowledge of disk sets.

Then I can recreate the disk sets again by first creating new local state
replicas.

Quote:> Having said that, you actually said:

> == IF you cant remove the metasets, then you can remove the "local set" and
> == reinitiate that, that should wipe any knwledge about the metasets from
> it.

> I actually originally read that you meant "remove the local state replicas".
> Meaning, I presume you meant "metadb -d", without specifying any diskset
> parameter, until all local state replicas are deleted, will effectively
> wipe out any knowledge of disk sets.

when you specify any meta command and dont give a -s <metaset> you are
working with the default metaset, if you clean that, i.e remove all
metadevices and metadatabases and start over, it should NOT remember that
you had such before.

Quote:> Then I can recreate the disk sets again by first creating new local state
> replicas.

/Johan A

        metaset -s testset -P

... and the diskset is now gone.

Interestingly, the -P option is not in the man page of metaset in Solaris9,
but it seems that it is in Solaris10. Fortunately, it also works on
Solaris9.