Decommission of datanode

[#1]

https://cwiki.apache.org/confluence/display/AMBARI...

When the REST API is used to trigger the decommission of the datanode a request ID is returned since its an async operation (see https://github.com/apache/ambari/blob/trunk/ambari...)

After polling on the status of the returned request ID for sometime, the request status shows "COMPLETED". However the decommission is actually still progressing (incomplete!)

Is this a bug ? Shouldn't the Ambari asynchronous request be marked COMPLETED only after the decommission has finished completely ?

[#2 · Accepted solution]

The decommission request from does not wait for decommission to complete - it issues the decommission command to hdfs and then hdfs asynchronously continues the decommissioning operation. At this point the jmx data from NN has to be monitored from the outside to know when hdfs completes decommissioning.

[#3]

Thanks. It would be helpful to update the cwiki.apache.org URL referred above with information on how to wait for decommission to complete