On Oct 5, 3:24 am, Huge <H...@nowhere.much.invalid> wrote:
> On 2007-10-04, Michael Tosch <eed...@NO.eed.SPAM.ericsson.PLS.se> wrote:
> > 4.
> > metareplace -e mirror device
>
> So it's OK to metareplace a disk with itself (as it were...)?
>
> --
> "Religion poisons everything."
> [email me at huge {at} huge (dot) org <dot> uk]

Yes, it is. That is what '-e' does for you. I had the same question
the first time I did this. Had my fingers crossed, rabbit's foot on
hand, and tossed some pixie dust over my shoulder praying I wasn't
about to blow anything up.

Doesn't the SB2000 use FC disks? If so, you will have to contend with
the WWN change as well.

Mark

On 2007-10-07, Mark <md2600@gmail.com> wrote:
> On Oct 5, 3:24 am, Huge <H...@nowhere.much.invalid> wrote:
>> On 2007-10-04, Michael Tosch <eed...@NO.eed.SPAM.ericsson.PLS.se> wrote:
>> > 4.
>> > metareplace -e mirror device
>>
>> So it's OK to metareplace a disk with itself (as it were...)?
>>
>
> Yes, it is. That is what '-e' does for you.

Fab, thanks. (The replacement disk hasn't arrived yet.)

> I had the same question
> the first time I did this. Had my fingers crossed, rabbit's foot on
> hand, and tossed some pixie dust over my shoulder praying I wasn't
> about to blow anything up.
>
> Doesn't the SB2000 use FC disks?

Yes.

> If so, you will have to contend with
> the WWN change as well.

Could you elaborate? (I know what an WWN is, but I thought that if I powered
down and swapped the disk, providing it isn't a boot disk, I don't need to do
anything further? I'm not hotswapping, so I don't need to use luxadm to change
the WWN of the running disk, and it's not a boot disk, so I don't need to
change the hardware path in the OBP, so I should be OK, shouldn't I?)


--
"Religion poisons everything."
[email me at huge {at} huge (dot) org <dot> uk]

On Oct 8, 3:34 am, Huge <H...@nowhere.much.invalid> wrote:
> On 2007-10-07, Mark <md2...@gmail.com> wrote:
> > If so, you will have to contend with
> > the WWN change as well.
>
> Could you elaborate? (I know what an WWN is, but I thought that if I powered
> down and swapped the disk, providing it isn't a boot disk, I don't need to do
> anything further? I'm not hotswapping, so I don't need to use luxadm to change
> the WWN of the running disk, and it's not a boot disk, so I don't need to
> change the hardware path in the OBP, so I should be OK, shouldn't I?)
>
> --
> "Religion poisons everything."
> [email me at huge {at} huge (dot) org <dot> uk]

You may not need to do anything. As long as you confirm that the /dev/
[r]dsk entries are updated to point to the new /devices entries which
are WWN specific, then you're OK. Or more to the point, as long as you
can see the new disk in format, then you're OK. Otherwise, running
devfsadm should fix you up. If you have sunsolve access, refer to
infodoc 76438. That one is for FC based servers, but the techniques
should be the same. luxadm is just for hot-swapping the disk. You may
still need to run devfsadm. 'boot -r' should work too.

You do not need to metadetach the failed mirrors. You can boot with
just one disk as long as you have removed the metadbs from the failed
drive. So you can just:

1. Use metadb -f to force delete the metastate databases from the
failed drive.
2. Power down
3. swap the disk
4. boot -rs from the good drive
5. check format, deal with dev names if necessary
6. prtvtoc|fmthard (assumes the disks have the same geometry ... if
not may well fail in which case use format)
7. recreate you metadbs
8. metareplace -e

Something like that.

Cheers,
Mark

On 2007-10-08, Mark <md2600@gmail.com> wrote:
> On Oct 8, 3:34 am, Huge <H...@nowhere.much.invalid> wrote:
>> On 2007-10-07, Mark <md2...@gmail.com> wrote:
>> > If so, you will have to contend with
>> > the WWN change as well.
>>
>> Could you elaborate? (

[snippage]

> You may not need to do anything.

That sounds good to me ....

[snippage]

> You do not need to metadetach the failed mirrors. You can boot with
> just one disk as long as you have removed the metadbs from the failed
> drive. So you can just:
>
> 1. Use metadb -f to force delete the metastate databases from the
> failed drive.

Already had to do this. It wouldn't boot with metadb's on the failed disk.

> 2. Power down
> 3. swap the disk
> 4. boot -rs from the good drive
> 5. check format, deal with dev names if necessary
> 6. prtvtoc|fmthard (assumes the disks have the same geometry ... if
> not may well fail in which case use format)
> 7. recreate you metadbs
> 8. metareplace -e
>
> Something like that.

Smashing, thanks.


--
"Religion poisons everything."
[email me at huge {at} huge (dot) org <dot> uk]

On 2007-10-08, Mark <md2600@gmail.com> wrote:
> On Oct 8, 3:34 am, Huge <H...@nowhere.much.invalid> wrote:
>> On 2007-10-07, Mark <md2...@gmail.com> wrote:
>> > If so, you will have to contend with
>> > the WWN change as well.
>>
>> Could you elaborate? (

[snippage]

> You do not need to metadetach the failed mirrors. You can boot with
> just one disk as long as you have removed the metadbs from the failed
> drive. So you can just:
>
> 1. Use metadb -f to force delete the metastate databases from the
> failed drive.
> 2. Power down
> 3. swap the disk
> 4. boot -rs from the good drive
> 5. check format, deal with dev names if necessary
> 6. prtvtoc|fmthard (assumes the disks have the same geometry ... if
> not may well fail in which case use format)
> 7. recreate you metadbs
> 8. metareplace -e
>
> Something like that.

Indeed, almost exactly like that. It's resyncing as I type this. Thanks very
much for the reassurance. (And the Infodoc number. I got one of the sysadmins at
work to print it out for me.)

The only minor glitch was SVM bitching about "invalid relocation information"
for the failed disk (the DeviceID has the WWN in it) - metadevadm -r seemed to
sort that out, although I'm not entirely sure how, since I fiddled with it for a
while, got a lot of errors and only then noticed it was correct in the metastat.

BTW, who thought WWNs were a good idea? And why?


--
"Religion poisons everything."
[email me at huge {at} huge (dot) org <dot> uk]