When maintaining a Hadoop cluster, you will need to restart various service from time-to-time when/if you update Hadoop configurations.
I ran into a problem today with Ambari where I wanted to do a rollling restart of all of my DataNodes, but when I clicked on the “Restart DataNodes” entry in the “Restart” drop down the dialog indicated “There are no DataNodes to do rolling restarts”.
I was having some issues with the ambari-metrics family of services on a ‘pseudo-distributed’ cluster that I have installed on my workstation.
The symptoms were:
1. Ambari indicated the following CRITICAL errors in the Ambari Dashboard under the Ambari Metrics section
Connection failed: [Errno 111] Connection refused to rchapin-wrkstn:6188
2. After attempting to restart the ambari-metrics-collector via either the Ambari Dashboard or through the commandline (# ambari-metrics-collector [stop|start]) you see the following (similar) messages in the ambari-metrics-collector.log