I think we will fail a lot of cache tests if we do this. And I don't see any reason not to wait for exchange in case it's an initial topology start with no data.
But this does make sense for tests with changing topology, so probably we should add a boolean flag allowing to skip this step. -- Val On Tue, Mar 31, 2015 at 12:39 AM, Sergi Vladykin <[email protected]> wrote: > Guys, > > I noticed that we have awaitPartitionMapExchange() call in > startGridsMultithreaded() in tests. This way we are hiding from ourselves > whole class of repartitioning problems. I would prefer having this call per > test if needed and not have it by default. What do you think? > > Sergi >
