John, If you’re running a proxy in front of NiFi, you will definitely need to configure some sort of session stickiness. Otherwise you can run into other issues as well. For instance, some "asynchronous requests” like updating variables, etc. may intermittently fail because the request creates a “request object” in the background that is temporary and then polled by the UI, and that resource exists only on the node that the request was made to.
Thanks -Mark > On Aug 26, 2020, at 1:27 PM, jgunvaldson <jgunvald...@cox.net> wrote: > > Hi All, > > We use a API Manager from WSO2 to proxy our NIFI APIs. This provides > subscription, OAUTH and other features that are very useful > > Problem is when a node is disconnected in NIFI (don’t ask, happens more than > we would like) then our API Manager, having been configured to use that node > - will send a 500 back to the user. It is not advanced enough to try other > (possibly up) NIFI NODEs (has a round robin list). Next request gets usually > the next configured NODE.. > > Thus every x divided by the (number of nodes configured) request fails - when > x represents a disconnected NODE. > > Question. What are some of the options for hosting a “reliable” NIFI API in > respect to loss or disconnection of a NODE? > > Good question right? > > Best Regards, > John >
