1

where do people place their multi user meta store? I'm going to use mysql but I don't know were I should stick it. on the name node or on its own server?

2 Answers2

2

Hadoop can really trash the page buffers on a datanode so you probably will do well to keep mySQL off of any of them.

The recommendation to run Hive outside the cluster is a very good one, if only because you will have serious balance problems if you run it inside the cluster.

Ted Dunning
  • 306
  • 1
  • 6
0

It can be on any machine really.

We installed hive on a node that is not a cluster member (but is only used for staging data to load into hive) and setup the metastore mysql instance there.

Nathan Milford
  • 792
  • 2
  • 10
  • 21