If there is a failure during Pivotal HDB initialization while it's deployed using Ambari, on the Ambari dashboard UI you may see the status of HDB as Stopped, however if you try starting it, the start attempt will fail as it has not been initialized properly.
In order to reinitialize it, you will need to perform the below steps. However, please ensure that the error because of which the previous initialization attempt failed has been resolved.
Note: Logs to verify HDB initialization failure:
1. Ambari Dashboard UI shows the logs during initialization. 2. gpinitsystem logs under <$GPADMIN_HOME>/gpAdminLogs directory on the HDB master server
Once you have verified the cause of failure and have fixed it, you can proceed with the below steps.
1. Verify that all the HDB services are stopped. You can run the below command from HDB master server.
ps -ef | egrep silent
Note: Ideally there should be no HDB process running, and the output of above command should not yield anything. It may happen that all the stages of hdb initialization are not complete, but the current state has left hdb services running but unusable. Thus, you will have to stop the HDB services before proceeding ahead. You can use the below command to stop it
su - gpadmin ; source /usr/local/hawq/greenplum_path.sh ; gpstop -af
2. Identify the directories specified for the below parameters from the HDB configuration tab
- HDB master data directory - HDB segments data directory - HDB temporary directory
- HDB HDFS directory
3. Drop the directories identified by logging on to the segments and master server.
4. Once the directories have been dropped, from Ambari Dashboard UI start HDB services. When HDB starts, it verifies if the directories specified by the configured parameter exist, if not it goes ahead to initialize the database again.
If you have resolved the issue faced during previous failure, these steps will let you proceed.