[dm-devel] Read-Only devices, multipath and more

[Christophe Varoqui]

Le mardi 20 mai 2008 à 15:48 +0100, Alasdair G Kergon a écrit :
> On Fri, May 16, 2008 at 09:35:44PM +0200, christophe varoqui wrote:
> > fact: device-mapper refuse to load a RW multipath table on these paths
> > (ie without libdevmapper:set_task_set_ro())
> > question: should change this behaviour to allow to load a RW multipath
> > table on these paths and let the IO be failed by the storage hardware ?
> 
> Yes.  But relax this *only* for multipath targets, where the whole idea
> is to "hide" path problems from the layer above.

Fair enough.

> It would not make sense 
> in my view to do this for other DM targets.
>  

During my testing I remember seeing a RW linear map loaded over a RO
multipath map (through a genuine vgchange -ay, the PV being the
multipath).

> > Alasdair proposed to add more explicit table loading ioctl return code
> > when the failure is due to this ready-only paths issue (E_ROFS for
> > example).
> 
> I think this should be done too (for non-multipath and other errors).
> 
Right, I'm about the publish a changeset for supporting the whole RO
notion in the multipath-tools :
1/ printing the RO/RW information
2/ fallbacking to RO when RW loading fails
3/ new flag to force map reloading, which takes care of the RO->RW
promotion and RW->RO demotion scenarii

I'll adapt 2/ to your new return codes when available.


Thanks for caring,
cvaroqui