On Thu, Jul 01, 2010 at 03:52:31PM +0100, Luca Bigliardi wrote:
> On Thu, Jul 1, 2010 at 3:34 PM, Iustin Pop <[email protected]> wrote:
> 
> > Hrmm… That's a weird way to use the rpc interface. Why are you not
> > simply returning the string, and _Fail() in case you get a block device
> > error (in patch 7/12)?
> >
> > Note it is a valid way to use it, but in that case (since you're not
> > using the validation routines in rpc.py), you need to validate yourself
> > the tuple received, etc. In other words, what advantage do you get from
> > having two status fields?
> 
> Well I would like to be able to proceed if a node/net health problem occurs,
> but raise an error if the node is misconfigured (in this way a drbd module
> not loaded triggers the error).

I understand your point, but I'm not sure if this is a good idea. The
temporary failure could be hiding the fact that the node might be
healthy and have the wrong helper, etc.

I would rather say:
- any failure aborts the operation: per the other code, broken nodes
  should be offlined anyway (and yes, you need to skip an offline node)
- optionally, failures could be forced via the --force flag

iustin

Reply via email to