Very Computer

by **Kyle Tuck** » Tue, 24 Oct 2000 04:00:00

Hi,
I have 2 disks (c1t2d0 & c1t3d0), 3 slices on each (0,1,3), with
each corresponding slice (0/0, 1/1/ 3/3) making up a stripe metadevice. I
am mirroring each of these to another matched set of disks (c1t4d0 & c1t5d0).
Some SCSI errors occured yesterday to c1t4d0s0 and DiskSuite is reporting
that only that one slice needs maintenance. I would thing all metadevices on
that disk would be flagged as needing maintenance. Since they're not, and I
have no hot spare as this is a simple Kingston 4 disk silo, I have to figure
out how to replace the bad disk. Is my only way out to detach all three
mirrors and then rebuild the 3 stripes using c1t4d0 & c1t5d0? Here's the
metastat output. Release is Solaris 7 with SDS 4.2. Thanks much.

proteus# metastat d30
d30: Mirror
Submirror 0: d31
State: Okay
Submirror 1: d32
State: Needs maintenance
Pass: 1
Read option: roundrobin (default)
Write option: parallel (default)
Size: 209707008 blocks

d31: Submirror of d30
State: Okay
Size: 209707008 blocks
Stripe 0: (interlace: 32 blocks)
Device Start Block Dbase State Hot Spare
c1t2d0s0 0 No Okay
c1t3d0s0 10176 No Okay

d32: Submirror of d30
State: Needs maintenance
Invoke: metareplace d30 c1t4d0s0 <new device>
Size: 209707008 blocks
Stripe 0: (interlace: 32 blocks)
Device Start Block Dbase State Hot Spare
c1t4d0s0 0 No Maintenance
c1t5d0s0 10176 No Okay

--
- Kyle
----------------------------------------------------------------------
Kyle Tucker - Manager, Unix Systems Tel: (978) 816-0229
Proteome, Inc. Fax: (978) 922-3971

Beverly, MA 01915 http://www.proteome.com
----------------------------------------------------------------------

by **Raf LaPiet** » Wed, 25 Oct 2000 11:33:00

I believe the command is "metareplace -e". In your case it would be

metareplace -e d30 c1t4d0s0

-Raf

>Hi,
> I have 2 disks (c1t2d0 & c1t3d0), 3 slices on each (0,1,3), with
>each corresponding slice (0/0, 1/1/ 3/3) making up a stripe metadevice. I
>am mirroring each of these to another matched set of disks (c1t4d0 & c1t5d0).
>Some SCSI errors occured yesterday to c1t4d0s0 and DiskSuite is reporting
>that only that one slice needs maintenance. I would thing all metadevices on
>that disk would be flagged as needing maintenance. Since they're not, and I
>have no hot spare as this is a simple Kingston 4 disk silo, I have to figure
>out how to replace the bad disk. Is my only way out to detach all three
>mirrors and then rebuild the 3 stripes using c1t4d0 & c1t5d0? Here's the
>metastat output. Release is Solaris 7 with SDS 4.2. Thanks much.

>proteus# metastat d30
>d30: Mirror
> Submirror 0: d31
> State: Okay
> Submirror 1: d32
> State: Needs maintenance
> Pass: 1
> Read option: roundrobin (default)
> Write option: parallel (default)
> Size: 209707008 blocks

>d31: Submirror of d30
> State: Okay
> Size: 209707008 blocks
> Stripe 0: (interlace: 32 blocks)
> Device Start Block Dbase State Hot Spare
> c1t2d0s0 0 No Okay
> c1t3d0s0 10176 No Okay

>d32: Submirror of d30
> State: Needs maintenance
> Invoke: metareplace d30 c1t4d0s0 <new device>
> Size: 209707008 blocks
> Stripe 0: (interlace: 32 blocks)
> Device Start Block Dbase State Hot Spare
> c1t4d0s0 0 No Maintenance
> c1t5d0s0 10176 No Okay

>--
>- Kyle
>----------------------------------------------------------------------
>Kyle Tucker - Manager, Unix Systems Tel: (978) 816-0229
>Proteome, Inc. Fax: (978) 922-3971

>Beverly, MA 01915 http://www.proteome.com
>----------------------------------------------------------------------

by **Kyle Tuck** » Wed, 01 Nov 2000 02:58:03

Quote:> I believe the command is "metareplace -e". In your case it would be

> metareplace -e d30 c1t4d0s0

Well I had no hot spares as I pointed out so there was no metareplacing
to do. I ended up having to detach the each of the three submirrors and
clear the submirrors. I then put in the new disk, recreated the stiped
submirrors and resynced. It all went rather smoothly and without incident.
I have a newfound fondness for the simplicity of DiskSuite versus the
complexity of Veritas VM.

>>Hi,
>> I have 2 disks (c1t2d0 & c1t3d0), 3 slices on each (0,1,3), with
>>each corresponding slice (0/0, 1/1/ 3/3) making up a stripe metadevice. I
>>am mirroring each of these to another matched set of disks (c1t4d0 & c1t5d0).
>>Some SCSI errors occured yesterday to c1t4d0s0 and DiskSuite is reporting
>>that only that one slice needs maintenance. I would thing all metadevices on
>>that disk would be flagged as needing maintenance. Since they're not, and I
>>have no hot spare as this is a simple Kingston 4 disk silo, I have to figure
>>out how to replace the bad disk. Is my only way out to detach all three
>>mirrors and then rebuild the 3 stripes using c1t4d0 & c1t5d0? Here's the
>>metastat output. Release is Solaris 7 with SDS 4.2. Thanks much.

>>proteus# metastat d30
>>d30: Mirror
>> Submirror 0: d31
>> State: Okay
>> Submirror 1: d32
>> State: Needs maintenance
>> Pass: 1
>> Read option: roundrobin (default)
>> Write option: parallel (default)
>> Size: 209707008 blocks

>>d31: Submirror of d30
>> State: Okay
>> Size: 209707008 blocks
>> Stripe 0: (interlace: 32 blocks)
>> Device Start Block Dbase State Hot Spare
>> c1t2d0s0 0 No Okay
>> c1t3d0s0 10176 No Okay

>>d32: Submirror of d30
>> State: Needs maintenance
>> Invoke: metareplace d30 c1t4d0s0 <new device>
>> Size: 209707008 blocks
>> Stripe 0: (interlace: 32 blocks)
>> Device Start Block Dbase State Hot Spare
>> c1t4d0s0 0 No Maintenance
>> c1t5d0s0 10176 No Okay

--
- Kyle
----------------------------------------------------------------------
Kyle Tucker - Manager, Unix Systems Tel: (978) 816-0229
Proteome, Inc. Fax: (978) 922-3971

Beverly, MA 01915 http://www.proteome.com
----------------------------------------------------------------------

by **Darren Dunha** » Wed, 01 Nov 2000 09:31:02

> Well I had no hot spares as I pointed out so there was no metareplacing
> to do. I ended up having to detach the each of the three submirrors and
> clear the submirrors. I then put in the new disk, recreated the stiped
> submirrors and resynced. It all went rather smoothly and without incident.
> I have a newfound fondness for the simplicity of DiskSuite versus the
> complexity of Veritas VM.

You just said you had to detach 3 submirrors, remove the submirrors,
then recreate all the mirrors. I'm not saying that it's a difficult
task, but I fail to see how that's "simpler" than telling veritas to
remove a disk for replacement, and then that you're replacing a removed
disk.

In addition, did you have any metadb replicas on the disk? Did you have
to create new ones on the new disk?

--

Unix System Administrator Taos - The SysAdmin Company
Got some Dr Pepper? San Francisco, CA bay area
< Please move on, ...nothing to see here, please disperse >

by **Kyle Tuck** » Wed, 01 Nov 2000 10:19:41

Quote:
>> Well I had no hot spares as I pointed out so there was no metareplacing
>> to do. I ended up having to detach the each of the three submirrors and
>> clear the submirrors. I then put in the new disk, recreated the stiped
>> submirrors and resynced. It all went rather smoothly and without incident.
>> I have a newfound fondness for the simplicity of DiskSuite versus the
>> complexity of Veritas VM.

> You just said you had to detach 3 submirrors, remove the submirrors,
> then recreate all the mirrors. I'm not saying that it's a difficult
> task, but I fail to see how that's "simpler" than telling veritas to
> remove a disk for replacement, and then that you're replacing a removed
> disk.

The number of steps may be more, but in my mind, the underlying concepts
of SDS and the parts that make up metadevices are more straightforward and
cleaner than VM's plexes and the like, particularily once you start building
logical volumes with parts from any number of places in the disk groups.
Also, had something gone wrong in the rebuild process, I could have easily
reverting to using the raw devices whereas I don't what you could do - if
anything - with Veritas to go back to square 1.

Quote:> In addition, did you have any metadb replicas on the disk? Did you have
> to create new ones on the new disk?

No, I had 4 metadb's on slices on my two internal OS disks on a small slice
5 (32MB).

--
- Kyle
---------------------------------------------

UNIX Sys Admin "Grandpa, what was NT?"
---------------------------------------------

by **Yousuf Kha** » Wed, 01 Nov 2000 13:55:56

Quote:> Well I had no hot spares as I pointed out so there was no
metareplacing
> to do. I ended up having to detach the each of the three submirrors
and
> clear the submirrors. I then put in the new disk, recreated the
stiped
> submirrors and resynced. It all went rather smoothly and without

incident.

Well, I realize it's a moot point now, since you've replace it already,
but are you sure there was anything wrong with that disk at all? When
only one slice is bad, and the rest of the slices on the same disk are
fine, my feeling is that it was simply a SCSI hiccup. In cases like
this in the past, I've simply just reactivated that disk, resynced all
of its data and it continued to work for years afterwards.

Quote:> I have a newfound fondness for the simplicity of DiskSuite versus the
> complexity of Veritas VM.

In my opinion, nothing beats DS for mirroring boot disks and other
internal disks. You can go straight back to using the simple devices
very quickly. The VM encapsulation is on the other hand very pokey, it
is very finicky about what it considers a standard boot disk slice (for
example, many people put a /opt partition on their boot disks, but VM
can't handle that, and converts /opt into a volume outside of rootdg),
and it's very hard to go back to the simple devices if anything goes
wrong unless you use the VM tools beforehand.

I've been able to recover from DS boot disk disasters by simply booting
to CDROM and mounting the simple devices off the boot disk. That's a
little more difficult with VM boot disk disasters.

I'd say people should stick to VM for large storage arrays where
redundancy of the SCSI or FC paths are important, so they can use the
VM's DMP module. But leave the boot disk mirroring to Disksuite.

Sent via Deja.com http://www.deja.com/
Before you buy.

by **Thomas Tornblo** » Wed, 01 Nov 2000 18:52:54

> > I believe the command is "metareplace -e". In your case it would be

> > metareplace -e d30 c1t4d0s0

> Well I had no hot spares as I pointed out so there was no metareplacing
> to do. I ended up having to detach the each of the three submirrors and
> clear the submirrors. I then put in the new disk, recreated the stiped
> submirrors and resynced. It all went rather smoothly and without incident.
> I have a newfound fondness for the simplicity of DiskSuite versus the
> complexity of Veritas VM.

In case you have not hotspares, and you physically replace the broken
disk with a new at the same scsi-id, then the command really is
"metareplace -e", with the proper arguments. Notice that you only name
one component with "-e".

There is no need to detach all of the mirrors, and you will run with
less redundancy if you do.

"metareplace -e" is your friend also in case you have an intermittent
problem, like a PSU that fails but the disk is OK.

--

Sun Microsystems AB Fax: +46 8 631 1102

Very Computer

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay

DiskSuite - 1 stripe bad, others okay