Scenario 1:
The fsimage file and the HDFS editlog are written to several places, including mounting NFS.
A) NameNode Daemon Crash: Solution: Just restart the Namenode process
B) The host is omitted where the name Node is executed.
Decision:
- Run namenode on another host with empty dfs.name.dir
- Point dfs.name.dir to mount NFS, where we have a copy of the metadata. OR
- Use the --importCheckpoint option when starting namenode after specifying fs.checkpoint.dir in the checkpoint directory from Secondary NameNode
- Change the fs.default.name file to the URI of the backup host names and restart the cluster with all the subordinate IP address in the slave file.
Note. We can skip editing that might have happened after the last breakpoint.
Scenario 2:
Fsimage HDFS .
A) Node Daemon Crash: :
B) , Node.
:
- , dfs.name.dir (1)
- Namenode -importCheckpoint fs.checkpoint.dir Secondary NameNode
- fs.default.name URI IP- .
, , .
, , .