On Fri, Oct 05, 2012 at 04:20:21AM +0200, Michael Hanselmann wrote:
> 2012/10/5 Iustin Pop <[email protected]>:
> > On Fri, Oct 05, 2012 at 03:55:38AM +0200, Michael Hanselmann wrote:
> >> The check for file consistency didn't properly handle virtual paths
> >> in case of a virtual cluster. This didn't cause any breakage as in
> >> a standard virtual cluster setup with only one node all files are
> >> visible for every node.
> >
> > LGTM, thanks.
> >
> > Do I understand correctly that what we're sending over RPC now are
> > virtualised paths? If so, do we have a check/list of what RPCs have
> > changed from file paths to the new virtual paths (all?)?
> 
> We don't, and I'm somewhat afraid there might be more. It isn't as
> easy as changing the RPC encoding either as some are hidden in data
> structures. File storage paths in logical IDs are still left
> untouched. Do you have a better idea for what we could do?

Nope, sorry. I was mainly asking because I need to keep this is mind,
and introduce a new type, when/if we add support for path-containing
RPCs in Haskell (at least there the separate type will solve the issue
of correctly virtualising all paths, but this doesn't help the Python
side).

> I will not yet push this patch as I noticed /etc/hosts is special (not
> virtualized). While virtual clusters might not want to modify
> /etc/hosts in the first place, an exception shouldn't be thrown on
> virtualizing paths if “--no-etc-hosts” wasn't specific while
> initializing the cluster.

Ack.

thanks,
iustin

Reply via email to