Hello,

Solaris 2.6 on a Sun UE450.

Yesterday, we had a big power loss here and the UPS/Generator combo did not
work. All the servers loss power. We're not a 24/7 shop, so I was called
upon to restart the servers.

We had a power supply on another machine that went berzerk/toasted (I
presume there has been a bad power spike) and one of the mirrored boot disk
on one of my sun box probably went bad in some way as well.

At the "ok" prompt, after typing "boot", the system would begin to boot and
30 second later come up with a message like: Short read 0x2000 chars read ,
disk error, cannot load /drv/.... (don't recall the end of the message
totally)

I then realized that the proper devalias for the mirrored boot drive was not
right in nvram, so I poceeded to redefine the mirrored boot drive correctly
and it successfully booted.

What is strange now, is that a metastat d110 (D110 being my mirrored boto
disk) comes up with no drive in "maintenance" mode. I don't think I can
trust that disk anymore, can I ? If I decide to replace itm how do I proceed
?

If it were in "needs maintenance" mode I would probably simply have to
hot-plug remove it, put a replacement disk, label and partition it, and then
use metareplace -e ...

But the disk and metadb does not seem to exhibit errors ! Why it did not
boot, I wonder, then !

If I decide I don't trust that disk anymore, do I simply have to metadettach
the three partitions on which (/, SWAP, /export/home) are mounted that are
on that are mirrored on that disk, metaclear the underlying concat/stripes,
physically replace the failed drive and simply recreate concat/strips
metadevice and reattach them to the one way mirror ?

Thanks for your advice

G Dahler wrote:
[..]
>
> If I decide I don't trust that disk anymore, do I simply have to metadettach
> the three partitions on which (/, SWAP, /export/home) are mounted that are
> on that are mirrored on that disk, metaclear the underlying concat/stripes,
> physically replace the failed drive and simply recreate concat/strips
> metadevice and reattach them to the one way mirror ?
>
> Thanks for your advice
>
I suppose you are running disk suite 4.2.1. The docs provide the answert
to the question how to replace your broken submirror:
http://docs.sun.com/app/docs/doc/806...tasksnew-17368

Concerning the short read - how about running an analyze with the format
utility before dumping the disk into the trash?

Tom

"G Dahler" <gd-nntp3> writes:
>
> If it were in "needs maintenance" mode I would probably simply have to
> hot-plug remove it, put a replacement disk, label and partition it, and then
> use metareplace -e ...
>
> But the disk and metadb does not seem to exhibit errors ! Why it did not
> boot, I wonder, then !
>

I wouldn't trust it either- it might be some kind of creeping failure.
Its happened to a couple drives I've seen over the last 10 years- they
work with the occasional quirks, but sooner or later they start really
failing. I'd get a replacement disk in there asap just on general
principles and put this one in a non-production array someplace where
you can mess with it.

Gregm

Thanks to all who responded.

I have called support and ordered a replacement disk which I just installed.

Before replacing it, I removed the mirrored slice and did an format/analyze
on the disk. Strangely, it did not report any errors. But the disk would
not boot yesterday, even after multiple tries. I still wonder what might
have happened. I replaced it anyway and it is now syncing....

Thanks