Greetings all. I've been looking for documentation to answer a couple of hadoop configuration questions but so far have come up dry. I was hoping someone could provide a quick pointer or answer a quick question(s).
We've had trouble with name resolution affecting hadoop (0.20 on top of which we run hbase), and have mostly made these issues go away by paring down the configuration down, eliminating aliases, and making dns match forward and reverse lookups to hostname. This seems to have resolved most of the issues, but one question remains. In hadoop configuration there is a configuration item, "dfs.datanode.dns.inerface" which takes the name of an interface. It seems this is not required to be configured and would be irrelevant on a single-homed server. In the case where you have an external and internal interface, like Amazon or Rackspace type environments, and your dns servers in resolv.conf are routed through one or the other, why would you need this parameter? Normal tcp/ip routing should handle that should it now? Is it necessary and does it have any benefit or impact on hbase? Thanks Greg
