Re: [gentoo-user] Weird "Buffer I/O error" in dev ndb0

[Holly Bostick]

Richard Fish schreef:
> Phill MV wrote:
>
>> Well, *something* is trying to access a function that doesnt exist
>> in NDB; whether we know what it does or why, I'm guessing it's a
>> behaviour that shouldn't happen :P.
>>
>> I suppose whatever tries to access NDB has a bug in it... but what
>> would that be?
>
>
>
> Do you have lvm installed (/sbin/lvm)?
>
> If so, the Gentoo startup script /sbin/rc will try to execute it
> after starting udev to probe for physical volumes, volume groups, and
> logical volumes. The default configuration of lvm probes _all_ block
> devices, so this is a probable suspect.  You can try adding a filter
> to /etc/lvm/lvm.conf in this case, like so:
>
> filter = [ "r|/dev/nbd.*|" ]
>
> As for whether it should or shouldn't happen, I am not sure.  I don't
>  know how common it is to use network block devices...it is
> definitely a high-end feature.  I would guess that anyone using nbd
> would also be interested in using lvm, so it is probably more useful
> to have lvm probe these devices if they exist.


Well, insofar as it is a confirmation of anything, I can say that that
explains the problem with nbd in my case; I have an lvm2 volume, managed
by evms (which is also too hige-end for my use, but I'm kinda stuck with
it now). Evms is also prone to probe every d*mn thing it can find, and
of course the lvm plugin is in use, because I actually do have an lvm2
volume. So I can see how if nbd was available, it would want to check
that as well, and both why removing it was necessary to make it stop, as
well as why it was unnecessary for me in the first place.

Thanks for the explanation.

Holly
-- 
gentoo-user@g.o mailing list