Hi!

I've got raid0 on my Sun StorEdge A1000 under Sparc Solaris 8.
I know, raid0 is not best solution :-(
Recently I was not able to mount my StorEdge.
I saw red error light on one of the disks.

Raid Manager software 6.22.01.14
shows error for that disk:
"A drive failed because it experienced a write failure"

And it shows that my raid0 is dead

I dismounted this failed driver from array
and try to read its data on PC.

AND I CAN READ IT! I saw hex data!

I mounted drive back into array and turn it on.
And I've got green light on failed disk!

But I still can't mount array.

Raid Manager shows that failed
disk is "replcaced", needs format and raid0 is still dead.

Can somebody help me with questions:

1. It's the same disk, why is my array dead?
2. Is it possible to recover my data?
(for example, using dd command to dump disk,
diskedit to edit it and so on)
3. May be I need to update service information on disks header or
I need to edit information in array controller memory?

Klim.

"Klim Samgin" <klimsamgin@yahoo.com> wrote in message
news:f5cc7e25.0312030634.3edaf7b9@posting.google.c om...
> Raid Manager software 6.22.01.14
> shows error for that disk:
> "A drive failed because it experienced a write failure"
>

If the controller cannot write to a destination drive (as the message
indicates), the drive is spun down. With RAID-0, this means at least one
block of data did not get written to disk. If the system crashed at this
point, probably many blocks did not get written to disk. If there was an
easy way to trick the volume back into existence, you would have no
confidence that whatever data remained, was intact, nor where any missing
blocks might be.

Your best bet is to delete the LUN, recreate it, reformat, reload from
tape...
Backups are always a requirement, but doubly so on RAID-0.

And I'd think real hard about re-using the drive that failed. I wouldn't
ever use it in another RAID-0 LUN, but I'm not sure I'd be too happy with it
in any other configuration, either.

"Kurt Duncan" <kurt.duncan@lsil.com> wrote in message news:<bqnluq$c0m$1@news.lsil.com>...
> "Klim Samgin" <klimsamgin@yahoo.com> wrote in message
> news:f5cc7e25.0312030634.3edaf7b9@posting.google.c om...
> > Raid Manager software 6.22.01.14
> > shows error for that disk:
> > "A drive failed because it experienced a write failure"
> >
>
> If the controller cannot write to a destination drive (as the message
> indicates), the drive is spun down. With RAID-0, this means at least one
> block of data did not get written to disk. If the system crashed at this
> point, probably many blocks did not get written to disk. If there was an
> easy way to trick the volume back into existence, you would have no
> confidence that whatever data remained, was intact, nor where any missing
> blocks might be.
>
> Your best bet is to delete the LUN, recreate it, reformat, reload from
> tape...
> Backups are always a requirement, but doubly so on RAID-0.
>
> And I'd think real hard about re-using the drive that failed. I wouldn't
> ever use it in another RAID-0 LUN, but I'm not sure I'd be too happy with it
> in any other configuration, either.

Hi!

Thank you for your answer.

I have halted Sun server with storedge in the evening
and I have turned them on in the morning.
The server can't mount storege than
and I saw this problem.
So, there were no writings when storedge crashes.

This storedge is using to store distributives, cdrom images and so on.
There were many readings, but few writings.
So, the storedge data was consistent at crash moment.

If I have a byte-to-byte copy of the failed drive,
from the same manufacturer, the same model and size, and so on,
why is array manager software marked it "replaced"?
I think, array manager software get this answer from storedge controller.
Why is storedge controller think so?
May be it checks raid disks headers, for example,
reads first sector of every disk, and looking for a label...

If so, I can mark every disk with same label
and I get recovered LUN.

What do you think about this possibility?

Klim.

Klim,

If I remember correctly this contains an old LSI
controller. These units store a DAC Store header
on disks in the array. This identifies what drives
are in what LUNs and what volumes are in those
LUNs.

If you replaced the drive then the system is recognizing
that this drive identifier is missing (similar to the
private region that Veritas Volume Manager uses).

I think you are better off reloading from tape.

Ken S.

"Klim Samgin" <klimsamgin@yahoo.com> wrote in message
news:f5cc7e25.0312050010.51b5b456@posting.google.c om
"Kurt Duncan" <kurt.duncan@lsil.com> wrote in message
news:<bqnluq$c0m$1@news.lsil.com>...
> "Klim Samgin" <klimsamgin@yahoo.com> wrote in message
> news:f5cc7e25.0312030634.3edaf7b9@posting.google.c om...
> > Raid Manager software 6.22.01.14
> > shows error for that disk:
> > "A drive failed because it experienced a write failure"
> >
>
> If the controller cannot write to a destination drive (as the message
> indicates), the drive is spun down. With RAID-0, this means at least one
> block of data did not get written to disk. If the system crashed at this
> point, probably many blocks did not get written to disk. If there was an
> easy way to trick the volume back into existence, you would have no
> confidence that whatever data remained, was intact, nor where any missing
> blocks might be.
>
> Your best bet is to delete the LUN, recreate it, reformat, reload from
> tape...
> Backups are always a requirement, but doubly so on RAID-0.
>
> And I'd think real hard about re-using the drive that failed. I wouldn't
> ever use it in another RAID-0 LUN, but I'm not sure I'd be too happy with
it
> in any other configuration, either.

Hi!

Thank you for your answer.

I have halted Sun server with storedge in the evening
and I have turned them on in the morning.
The server can't mount storege than
and I saw this problem.
So, there were no writings when storedge crashes.

This storedge is using to store distributives, cdrom images and so on.
There were many readings, but few writings.
So, the storedge data was consistent at crash moment.

If I have a byte-to-byte copy of the failed drive,
from the same manufacturer, the same model and size, and so on,
why is array manager software marked it "replaced"?
I think, array manager software get this answer from storedge controller.
Why is storedge controller think so?
May be it checks raid disks headers, for example,
reads first sector of every disk, and looking for a label...

If so, I can mark every disk with same label
and I get recovered LUN.

What do you think about this possibility?

Klim.