Step C: Split the cluster - AWS Elemental Conductor Live

Step C: Split the cluster

Note

In this procedure, the Conductor nodes change their roles. Therefore, the node that is originally the primary node is called node X. The node that is originally the secondary node is called node Y.

To split the cluster into two clusters, you first fail over control to the secondary Conductor node (node Y), then you remove the primary Conductor node (node X).

You start with the original deployment of one cluster, which looks like the following diagram.

Diagram showing original cluster with two conductor nodes and multiple worker nodes.

You end up with:

  • The original cluster controlled by the second node. This cluster is using the older version of the software. For example, Conductor Live 3.25.5 and Elemental Live 2.25.5.

  • A new cluster that right now consists only of the primary Conductor node. The Conductor is using 3.26.x.

During this time, the worker nodes (on the original cluster) continue processing without interruption.

  1. Disable high availability (HA) on the cluster. See Enabling or disabling high availability (HA).

  2. Create a backup of the data on node X. See Backing up data.

    Important

    Now that you have made a backup of this Conductor node, don't make any changes to this Conductor node or to cluster until you've finished this migration process. Don't change the setup of the Conductor node, don't create channels, don't create new node assignments for any channel, and so on.

  3. On the web interface for node X, re-enable HA. See Enabling or disabling high availability (HA).

  4. Perform the commands in the next few steps from the command line.

    1. From the Linux prompt, use the elemental user to start a remote terminal session with node X.

    2. Fail over to the secondary Conductor node. To fail over, enter these commands from the CLI on node X:

      sudo systemctl stop elemental_se
sudo systemctl stop serf

    Node Y is now controlling the cluster.

  5. To enable user authentication on node Y, enter this command from the CLI:

    sudo systemctl restart httpd

  6. On the web interface for node Y, disable HA. See Enabling or disabling high availability (HA).

  7. Now that HA is disabled, you can remove node X from the cluster. See Removing a Conductor node from the cluster.

  8. On the CLI, enter the following commands to push the VIP into node Y:

    sudo systemctl start keepalived
sudo systemctl enable keepalived

Node Y is now the Conductor that is controlling the cluster. Note that you no longer have Conductor redundancy in the cluster.

Diagram showing original cluster with Node Y as Conductor, Node X, and multiple worker nodes.