Github user danny0405 commented on a diff in the pull request:
https://github.com/apache/storm/pull/2319#discussion_r138091372
--- Diff: storm-core/src/clj/org/apache/storm/daemon/supervisor.clj ---
@@ -758,6 +752,35 @@
(catch Exception e
(log-error e "Error running profiler actions, will retry again
later")))))
+(defn assigned-assignments-to-local!
+ [^SupervisorAssignments supervisorAssignments supervisor]
+ (when (not-nil? supervisorAssignments)
+ (let [serialized-assignments (into {} (for [[tid amt]
(.get_storm_assignment supervisorAssignments)]
+ {tid (Utils/serialize amt)}))]
+ (.sync-remote-assignments! (:storm-cluster-state supervisor)
serialized-assignments))))
+
+;; Supervisor should be told that who is leader.
+;; Fetch leader info each time before request node assignment.
+;; TODO: get leader address from zk directly.
+(defn assignments-from-master
+ [conf supervisor]
+ (let [client (atom nil)]
+ (try
+ (let [master-client (NimbusClient/getConfiguredClientAs conf nil)
+ _ (reset! client master-client) ;; keep a refence so we can
close it
+ supervisor-assignments (.getSupervisorAssignments (.getClient
master-client) (:my-hostname supervisor))]
+ (assigned-assignments-to-local! supervisor-assignments supervisor))
+ (catch Throwable e
--- End diff --
okey, but if we just throw it out, supervisor will restart more frequently,
i have no idea now, this need a discussion
---
