Trivial-whether or not to set SSH password-free

 

Hadoop2.2.0 has been set up many times, and many different startup methods, configuration parameters, and locations have been tried, which involves whether to configure SSH password-free logon or not.

First, find out why SSH password-free login is required.

SSH password-free logon procedure:

......

There are three startup methods in hadoop2.2.0:

1. Start each daemon using the hadoop-daemon.sh and yarn-daemon.sh;
2. Start all processes of different modules separately using start-dfs.sh and start-yarn.sh;
3. Start all processes in all modules with a start-all.sh.

Three startup methods can be used in the experiment stage, but the first method is more suitable for the production environment.

 

The first type does not require SSH password-free login, but the second and third types do. Why?

The first is to start every process on each machine one by one. That is to say, in a completely distributed environment, how many machines are there? I will use this command on each machine to start the corresponding commands, for example, to start namenode on namenode, you must useSbin/hadoop-daemon.sh start namenodeThis command is used to start datanode on datanode.Sbin/hadoop-daemon.sh start datanodeThis command, the corresponding yarn process starts the same, but it uses the yaen-daemon.sh.

The second and third methods are convenient to start the entire cluster only when used on namenode. The principle is to start the corresponding process on namenode according to the configuration file and thenLog on to slave through SSH to start the corresponding process.

 

Now that I understand this, my previous saying or understanding is wrong.

Because you only need to remotely log on to the Server Load balancer instance to start the corresponding process, all Server Load balancer instances only need to have the public key of the namenode and do not need the public key of other Server Load balancer instances.

However, I guess that the namenode and slave information are read in the configuration file, so if slave can log on to namenode through SSH without a password, can I format the file system on namenode and start the cluster on a server Load balancer instance? Haha.

 

Do I need to use the SSH protocol elsewhere?

The answer is no. It is used only when it is started. (My current understanding)

Others, such as heartbeat, are used.RPC protocol, Web UI usedHTTP protocolIn hadoop, RPC and HTTP are much more important than ssh.

 

Trivial-whether or not to set SSH password-free