On Thursday, June 20, 2019 12:00:38 PM MSK Konstantin Osipov wrote:
> * Stanislav Zudin <szudin@tarantool.org> [19/06/19 15:53]:
> > Do not spread the space:truncate() to replicas if the
> > affected space is local and temporary.
> 
> While I agree with Georgy's comment that when it comes to
> TEMPORARY spaces, we should replicate all DDL but TRUNCATE, for
> node-local spaces I think nothing should be replicated. This is
> the idea of node-local spaces to begin with.

It breaks backward compatibility and inconsistent all. I'm pretty sure we 
definitely should replicate all local space schema changes because it is only 
possible way to keep a fallback possible. In the opposite case we were in duty 
to keep all local spaces definition in mind and apply them before replica 
became master. The other solution is to populate local space changes across 
whole cluster by hands instead of replication and it looks as ugly as previous 
one.