RU is meant to have as little downtime is possible. In that case, we bring up a new instance on a new port and issue a --deregister the old version. Doing so will not accept new requests on the "old" instance. Once all requests are finished then the "old" process will close on it's own while the "new" process services clients.
For EU, there is inherent downtime, so there's no need to change the port. Thanks On 1/28/16, 3:52 PM, "Di Li" <[email protected]> wrote: >Hello folks, > >I noticed that in rolling upgrade Ambari changes HiveServer2 port to >10010 >(or 10011, depending on the transport mode). Ambari does not however >change HiveServer2 port in express upgrade. >I dug out the JIRA AMBARI-9420 where this RU behavior was added, but it >doesn't explain why it had to be done so. > >Could someone help explain why this RU-only behavior ? > >Thank you. > >Di Li >InfoSphere BigInsights >IBM Ottawa Lab Palladium Campus >Email: [email protected] >Phone: 613-270-4883, Tie-line:315-4883 >
