[Yaird-devel] Re: evms first try

Erik van Konijnenburg ekonijn at xs4all.nl
Sun Dec 4 22:31:04 UTC 2005


On Sun, Dec 04, 2005 at 05:14:19PM +0100, Marco Amadori wrote:
> Alle 19:06, sabato 3 dicembre 2005, Erik van Konijnenburg ha scritto:
> > On Fri, Dec 02, 2005 at 11:56:26AM +0100, Marco Amadori wrote:
> > > Alle 00:36, venerdì 2 dicembre 2005, hai scritto:

> meanwhile, this patch makes yaird fail later...
> ...
> with "yaird error: No block special file for dm-4 in /dev/mapper or /dev/evms 
> (fatal)"
> 
> this is strange since dm-4 is not a evms_volume nor an evms disk.
> 
> So maybe I just do not understand what yaird tries to do. At this moment our 
> evms implementation scan only for evms volumes, an evms volume is an object 
> directly mountable or mkfs-able. In this previous case dm-4 was an evms 
> object but not an evms volume.
> 
> So I think I just need to make EvmsTab scans for all objects and not only for 
> volumes, right?
> 
> I'll try to do that.

I get the impression that the idea in EVMS is that you're only supposed
to be able to mount volumes, not objects; presumably that's why they hide
everything but volumes in /dev/evms/.nodes/.

Of course we could try to ignore the 'mount only volumes' idea, but then
you're working against EVMS concepts, not with it; I'm not in favour of
this approach.   At least not while there are other bugs that get in the way
of finding out whether ignoring it is really necessary.

> ...
> > Note that the above is as it should be, not as it currently is implemented.
> What are the differences? I saw in the code that it tryes to work like this.

In the previous version, yspecial made a difference between dm devices and other devices
but it did not distinguish whether the dm device is created by evms or lvm.

In Revision 17 (Sun 2005-12-04 22:51:15 +0100) this should be cleaned up;
if you could test that, great.

Regards,
Erik



More information about the Yaird-devel mailing list