Our product provisions clusters in an automated way using Ambari. About 1 in 50 clusters gets this error, so we're not sure exactly how to reproduce it. It might be a race condition of some sort. One of the first things we do is run a set of custom actions on all the nodes in the cluster. We do that by POST'ing a request to the cluster. Randomly that request will throw a 500 error with this NullPointerException:
https://gist.githubusercontent.com/jimbobhickville/62176c2053827a90efab/raw/34fec22ff0fda2056090377ca432b18f58073d9a/gistfile1.txt This is on Ambari 2.1.1 The code path looks pretty normal, but I'm not much of a Java dev so there could be something that isn't obvious to me. It's just looking up the current cluster in the database by name. Doing a GET /clusters/<cluster-id> works fine at the same point in the process, afaict, so I don't see how it's getting a NullPointerException when doing it internally. Is this a known issue? Is there a workaround or way to tell when it's safe to issue the request? Greg
