I ran a API call to decommsion datanode.   HDFS.decommission(dn.name) I used the following to make sure the role is indeed decomissioned: if role.commissionState == 'DECOMMISSIONED': then I will run role deletion: HDFS.delete_role(dn.name) But I got the error:   Removing datanode roles..Failed to remove datanode role on the host Role hdfs-DATANODE-4a8948d61dc8a4727f810f736d9d3447 has 1 active commands (error 400)   It turned out that after commissionState became DECOMMISSONED, the UI still shows the decommisiong command was running for another 10 to 15 seconds.   To workaround that, I had to let the prgram sleep for addtional 60 seconds after the decomission status became decomissioned.   Is this a known issue in API v11? ... View more

It makes perfect sense if we enable auto failover in a NN HA configuration. But if I just want to enable NN HA without enablhing auto failover, does it provide any advantage comparing single namenode + secondary namnode?   NOTE: in my non-ha configuration, I have mounted one of my nn dir to NFS by which provides fault tolerance. ... View more

According to the following github link, the pools_refresh() was availalbe since API V6. https://github.com/cloudera/cm_api/blob/master/python/src/cm_api/endpoints/clusters.py   When I check the source code of V11 clusters.py, I couldn't find this funciton, was it removed? ... View more

I am not 100% sure how an active namenode write to Jounral nodes. Considering the folloiwng example:   2 namenodes(active+standby) 3 journal nodes   When the active namenode updates edit file, does it write 1/3 copies of the updated edits to each jn? or it write a complete copy to each jn?   When standby nn read the edites from jn, does it read 1/3 copy from each jn or the entire copy from each jn? If it's later, what if the edit md5 doesn't match across the jn nodes?     Thanks! ... View more

I tried to use API.create_host(hadoop_roles.HOSTNAME, hadoop_roles.HOSTNAME, IP)  But when I tried to apply host template after this, it compained the parcel is not installed. I then tried execute parcel.start_distribution() parcel.activate() But the same problem remains.   So what is the API call to add a new host.   This problem is related to https://community.cloudera.com/t5/Cloudera-Manager-Installation/How-to-use-CM-API-to-distribute-parcel-to-a-host/m-p/45650#M8259 ... View more