Hi Clay, RPG (Site-to-Site) is a Peer-to-Peer communication protocol. There's no distinction between a primary node and the remote cluster, or nodes other than the primary node and the remote cluster. E.g. With Cluster A (node a1, a2 and a3) and Cluster B (node b1, b2 and b3) Each node must be able to communicate with every remote node. Node a1 will communicate with all of node b1, b2 and b3. So do node a2 and a3. Those connections are persistent. S2S RAW uses socket based connection. S2S HTTP uses PoolingHttpClientConnectionManager internally to reuse connection.
PostHTTP uses PoolingHttpClientConnectionManager, too. InvokeHTTP uses different HTTP client library, that is okhttp. I didn't check it, but I assume it supports keep-alive. Do you have any specific concern about keep-alive? The keep-alive technology can be used to improve performance. However, we should not depend on that for any load-balancing rule. If you are looking for a solution to distribute FlowFiles based on some rules, NIFI-5516 will be useful (under development). https://issues.apache.org/jira/browse/NIFI-5516 Hope this helps. Thanks, Koji On Tue, Sep 25, 2018 at 12:07 PM Clay Teahouse <[email protected]> wrote: > > Hi All, > > Are the connections between the primary node and RPG persistent, and if no, > is there a way to make them persistent? > > Similarly, are the http connection from PostHTTP and InvokeHTTP to the > destination persistent, meaning keep-alive is set to true? > > thanks a lot > > Clay
