Do you mean the connection between the Hive client and the Hive metastore (if you are using the command line?) or the connection between the metastore server code and the RDBMS. The connection to the RDBMS uses JDBC connection pooling to avoid making and tearing down many connections. The connection of the command line to the client to the Hive metastore uses thrift, which I believe by default builds a new TCP connection for each operation.
Alan. > On Dec 13, 2016, at 10:07 PM, Huang Meilong <ims...@outlook.com> wrote: > > Hi all, > > Will HMS keep the connection to meta db when HMS is up? Or will HMS build > connection to meta db every time the query comes to HMS and release > connection to meta db when query finished?
