Github user zd-project commented on a diff in the pull request:
https://github.com/apache/storm/pull/2789#discussion_r208353770
--- Diff:
storm-server/src/main/java/org/apache/storm/daemon/nimbus/Nimbus.java ---
@@ -2807,7 +2808,14 @@ public void launchServer() throws Exception {
}
});
- StormMetricsRegistry.registerGauge("nimbus:num-supervisors",
() -> state.supervisors(null).size());
+ StormMetricsRegistry.registerGauge("nimbus:num-supervisors",
() -> {
+ try {
+ return state.supervisors(null).size();
+ } catch (Exception e) {
+ e.printStackTrace();
--- End diff --
It's a band-aid fix for now to unblock flushing. I should probably redirect
e.printStackTrace() to LOG.error. To really fix the issue, we either have to
unregister this gauge manually, which is not ideal, or come up a way to notify
Nimbus of the shutdown of Zookeeper. This again relates back to the issue that
I brought up earlier about the connection refused exception from Nimbus to
Zookeeper, which I don't think have a good fix so far.
---
