Github user agresch commented on the issue: https://github.com/apache/storm/pull/2504 This JIRA currently only addresses metrics on the Nimbus leader. If Nimbus goes does down and switches to the standby, the options would be to sync the RocksDB database from one server to another, send metrics from the supervisors to both Nimbi (is that the correct pluralism?), or allow metrics to be dropped. We plan further work to HBase as an option as well, which would limit the metric loss here. I'd recommend dealing with Nimbus standby options in a future JIRA, but welcome discussion.
---