Sorry for the likely basic question, but I am having a little trouble making sense of the documentation. We are currently running a HDP 2.4 cluster, hosting an unsecured HBase database. All access to HBase is via Phoenix. We connect to Phoenix from our application using the fat client. So, we have no authorization/authentication/etc. We are now looking to start adding auth, and I want to make sure we are looking at the correct tools. We would like the ability to pass a user/password along when establishing our Phoenix connections, so we can begin to provide read-only access to the cluster to some users, and read/write access for others. Do we need to use Kerberos? We are not currently running this in our setup, and would like to avoid if possible - according to our security folks. No external customers will access our cluster, all access is via our app servers, or for engineers running ad-hoc queries for debugging purposes. It sounds like Ranger would be the tool to handle authorization (via Kerberos) and authentication for HBase. Does that work for Phoenix as well? Is PQS required, or would it work with the fat client as well? Is there something else we should be looking at, maybe a simpler way to create read-only users in our cluster? We are looking at upgrading to the latest stable HDP in the next several months, if that changes anything. ... View more

We are running HDP 2.4. Our Java application connects to Phoenix using the fat client. While testing our queries in a stand alone query tool (DbVisualizer), some queries will error out when we allocate 4GB to the client, and then work when we up that to 6GB. We know that there may be things we can do to reduce the memory usage of those queries, and we are working on that, but when we move to embed these Phoenix queries within our regular Java app, we need to make sure that a bad Phoenix query does not use up all of our heap and cause problems for the rest of our application. Aside from query timeouts, are there any settings we can enable to limit the amount of memory that the client will use? ... View more

I am trying to determine the best practices for using the IN clause in Phoenix. Is there a practical (if not physical) limit to how many vales we should include? Would 1000 elements be OK? If we want to pull back result sets which match a couple thousand keys, we planned on batching the keys into sets of 1000 on the application side and passing that to Phoenix, but want to make sure we are not going to get bit. Running HDP 2.4 ... View more