1) HDFs reading process
Distributedfilesystem, Fsdatainputstream, Dfsclient.open (RPC communication mechanism), Nn.open HDFs Write Process Distributedfilesystem, Fsdataoutputstream, dfsclient.create (RPC communication mechanism), nn.create 2) The function and mechanism of Secondarynamenode Secondarynamenode is not a complete sense of namenode a backup pull fsimage and edits files are merged in SNN memory Fs.checkpoint.periodfs.checkp Oint.sizefs.checkpoint.dir checkpoint node (same as Secondarynamenode, Secondarynamenode in HADOOP2 has been replaced by this) backup node Namenode Backup in full sense 3) once the Namenode or metadata information is lost, we can reply to our metadata information via the Secondarynamenode checkpoint directory Hadoop Namenode- importcheckpointhadoop-daemon.sh start Namenode 4) Rack-aware by default all Datanode are considered to be in the same rack, regardless of whether they are physically in a rack/ Default-rack If you need to configure a true network topology, you need to configure the following files: The Topology.script.file.name property value is a script that is written in the true sense of the network topology/D1/RECK1/DN1
HDFs Detailed Analysis II