@Michael Bronson
I guess this is what we should ideally do.
1. Check the "/var/log/ambari-server/ambari-server.log" to findout why the blueprint installation failed. In some cases may be checking the failed component logs also can help to know if any important configuration parameter is missing for that component in blueprint.
2. Based on above point we will need to fix the blueprint Or hostmapping.
3. Now before attempting to reinstall the blueprint this time better to do "ambari-server reset" so that the incomplete cluster information present inside the ambari server DB is cleaned.
4. While deploying blueprint based cluster perviously if some of the agents have started some processes like NameNode/DataNode/NodeManager ...etc then we should kill them before attempting to recreate the cluster using blueprint. Else there may be proess conflict when ambari will attempt to register and install the same components on those hosts again.
Based on our need we can set the "provision_action" inside the blueprint as "INSTALL_ONLY" so that ambari will just install the components and will not attempt to start them immediately which saves some time in case of failure as it does not wait long for components to be restarted. We can start the components later from ambari UI when the cluster is installed properly.
"provision_action" : "(INSTALL_AND_START | INSTALL_ONLY)"
provision_action : Cluster wide provision action can be specified in Cluster Creation Template (see below), but optionally this can be overwritten on component level, by specifying a different provision_action here. The default provision_action is INSTALL_AND_START. See: https://cwiki.apache.org/confluence/display/AMBARI/Blueprints
.
