1) Restart HMaster processes in a rolling fashion
Warning: Make sure there are several HMasters deployed in the cluster.
The following steps assume there are two HMasters in the cluster.
- Stop active HMaster (for example, HMaster1).
- Confirm if the other HMaster (for example, HMaster2) is in active mode.
- Start HMaster1 and confirm it is in standby mode.
- Stop HMaster2 and confirm HMaster1 is in active mode.
- Start HMaster2 and confirm it is in standby mode.
2) Sideline the obsolete WALs
Warning: Before proceed the instruction, report to Cloudera support to make sure the WALs can be sidelined.
(Commands need to be ran as hbase user or principal )
- Make a temporary directory (for example, hdfs dfs -mkdir /tmp/old-wals)
- Move the obsolete WAL files (for example, hdfs dfs -mv /hbase/WALs/<OBSOLETE-WALs> /tmp/old-wals/)
- Restart both HMasters in a rolling fashion