This section describes the operational changes that can occur when a node in an Avi Load Balancer controller cluster reboots.

Follower Reboots

Reboot of a follower node does not significantly impact the operation of the Avi Load Balancer Service Engines (SEs) connected to the node.

  • Logs are saved locally on the SE until connectivity to the follower node is restored.

  • After reboot, the controller node is configured to handle the virtual service logs from the SE.

  • Analytics to the node is briefly interrupted.

  • The SE reconnects over SSH when the node is back up.

  • The follower replicates the PostgreSQL database of the leader.

  • The web interface, REST API, CLI, and other controller services continue to be supported.

Related Event Messages

The following event messages are generated:

  • CONTROLLER_NODE_LEFT - when a follower node leaves the cluster.

  • CONTROLLER_NODE_JOINED - when a follower node rejoins the cluster.

Leader Reboots

When the leader node in an Avi Load Balancer controller cluster reboots, one of the follower nodes takes over as the new leader. Operation of the SEs connected to the node is not significantly impacted.

The process is similar to when a follower reboots. Additionally, the leader IP address is updated to that of the follower that has become the new leader.

Related Event Messages

The following event messages are generated:

  • CONTROLLER_NODE_LEFT: Generated when a follower node leaves the cluster.

  • CONTROLLER_NODE_JOINED: Generated when a follower node rejoins the cluster.

Note:

These are the same events generated when a follower reboots and rejoins.

Multiple Followers Reboot

The process of rebooting for multiple followers is similar to when a single follower reboots.

During the reboot process, the cluster status changes as follows:

  • Ready

  • Not Responsive

  • Restoring

  • Waiting for Service Engines

  • Cluster copying DB from leader

  • Ready

Related Event Messages

The following event messages are generated:

  • CONTROLLER_NODE_LEFT: Generated when a follower node leaves the cluster.

  • CONTROLLER_NODE_JOINED: Generated when a follower node rejoins the cluster.

Note:

These are the same events generated when a single follower reboots and rejoins.

Leader and Followers Reboot

When all the Avi Load Balancer controllers in the cluster reboot, the process is similar to the process when the leader and only one of the followers reboots.

During the reboot process, the cluster status changes as follows:

  • Ready

  • Not Responsive

  • Errors (if any)

  • Restoring

  • Waiting for Service Engines

  • Cluster copying DB from leader

  • Ready

Related Event Messages

The following event messages are generated:

  • CONTROLLER_LEADER_FAILOVER: Generated when a follower node leaves the cluster.

  • CONTROLLER_NODE_JOINED: Generated when a follower node rejoins the cluster.

The CONTROLLER_LEADER_FAILOVER indicates that a new leader has taken over. The CONTROLLER_NODE_JOINED event indicates when followers join the cluster.

Service Engine State Changes

The SE state changes as follows:

  • Avi Load Balancer Controller heartbeat failed.

  • Headless (SE is continuing to operate but without an Avi Load Balancer Controller.)

Once the cluster is back, the SE reconnects to the Avi Load Balancer Controller.