Configure User Accounts

Before starting the installation, identify the nodes that will participate in the solution. For a standard installation, include an edge (gateway) node, a node to host the UI (operational analytics), and one or more nodes to host the machine learning component. You also need to identify a user account to run the solution with. The recommended approach is to use a non-administrator account with access to HDFS, and either the same local account for all nodes or an account authenticated with Kerberos.

Configure User Accounts

For each identified node, create the solution user. This will also create the /home/solution-user directory. Set the same password for each account.

Solution user needs to have sudo privileges.

sudo adduser <solution-user>
passwd <solution-user>      

For unattended execution of the ML pipeline, a public key authentication will be required. Log on to the first ML node (usually the lowest-numbered node). and create a private key for the solution user. Then you will need to copy those credentials to each node used for ML.

[soluser@node04] ssh-keygen -t rsa
[soluser@node04] ssh-copy-id soluser@node04
[soluser@node04] ssh-copy-id soluser@node05
…..
[soluser@node04] ssh-copy-id soluser@node15

The sample above assumes that the solution-user is “soluser” and there are 12 nodes used for ML. Now do the same for the UI node.

[soluser@node04] ssh-copy-id soluser@node03