Scenario is managing a large number of temporary/ephemeral clusters so comparing metrics across clusters is required, even those which are no longer online. Primary goal is continual improvement and identifying issues introduced over time (such as changes in the application or updates to HDP).
When Ambari used Ganglia we could have exported the metrics before cluster shutdown. Or updated the Ganglia client config to point to a shared Ganglia server, and not needed to update that server since Ganglia has "discovery".
the new features in Ambari 18.104.22.168 and 22.214.171.124 are selectable metric time ranges and export as CSV and JSON. I don't know of any other way except for exporting metrics to a file. Images attached
Thanks. That would only get each individual metric and the file would be a csv/json which doesn't appear that useful in other systems. Though it could be scripted around (a lot of work) to grab all metrics, parse, and convert to a format to be used elsewhere.
Ideal state would be the ability to export all metrics as RDD. Then you simply throw Graphite in front of it. RDD is the standard in most metrics collectors.