Marco Peereboom <slash <at> peereboom.us> writes:
> Upon reboot the mirror should be brought up with only the surviving
> member. If this isn't the case please show me a trace so that I can go
> fix that bug.
'trace' means what here? Yes, I unplugged a drive of a working mirror as I
wrote, halt, plugged it back with power off, rebooted, and had the described
state of fsck problems, fortunately telling me exactly about the files that
were available on A ('.pid') and B ('testo') only.
> Drive A is DEAD. Do not EVER use it again.
That's fine, except it was tried to re-insert it into the mirror at reboot.
Not on my account, but on its own, as one might say.
> YOU CAN NEVER USE THE UNPLUGGED DRIVE EVER EVER EVER EVER EVER
> AGAIN!!!!! IT IS DEAD AND IS CORRUPT AND PUPPIES DIE WHEN YOU USE IT!!
>
> Hope this sinks in.
Fine. (Should I really say, it required some file checks first, and then,
and now, it automagically is a member again of my - by now - online raid?
Without me adding it?
> If the dead drive becomes a participating member of a raid set
> something is broken; very badly broken. Show me a trace, including
> bioctl output, if this is the case.
Again, bioctl I do understand. 'trace' only at panic.
> > Does this make sense?
>
> Not sure. I have a hard time following what you are doing vs. what your
> expectations are.
Then keep asking!
I do have the impression, what I wanted, is what you already had in mind:
a broken mirror simply remains dead and broken, and the machine runs happily
before and after reboot on the sane drive. Correct?
Uwe
