Currently Kudu servers have the ability to periodically dump metrics to the log.
KUDU-2279 improved the compactness and performance of the log so it can be enabled by default.
In addition to metrics, though, it would be useful to have some other machine-readable data periodically recorded by servers. For example, periodic stack traces or RPC traces could be helpful for understanding latency issues after-the-fact. This JIRA tracks the effort to convert the metrics log to a more general-purpose diagnostics log.