Failover Scenarios

When an Active node fails for any reason, the following failover scenarios are supported:

•Automatic failover, see Working with Automatic Failover.

This scenario requires specific configuration, see Prerequisites and Setting Up an Active/Passive Cluster.

To recover after an automatic failover, see Restoring a Cluster after an Automatic Failover.

•Manual failover using a dual-resolving DNS entry, see Performing a Manual Failover with a Dual-Resolving DNS Entry.

•Manual failover using a single DNS entry, see Performing a Manual Failover Using a Single DNS Entry.

•Manual failover using an IP address swap, see Performing a Manual Failover Using Interface IP Swap.

Working with Automatic Failover

Automatic failover of a two-node Active/Passive cluster is supported by Pulse One appliance.

When the Active node (A) fails, no user action is required to perform the failover itself:

•In a short time, the Passive node (B) detects the failure.

•The Passive node (B) makes itself a standalone node automatically. This node is effectively a standalone “Active” node which is based on the configuration of the original Active node (A). However, there is no longer an operational cluster.

•All PCS/PPS appliances that are at v9.0R2 (or later) adjust automatically to communicate with the new standalone node (B).

•Pulse One operations continue against the standalone node (B).

When the Passive node in a cluster fails, no action is performed automatically by Pulse One. The Active nodes is unaffected, and the cluster persists when the Passive node returns to service.

To recover the cluster, see Restoring a Cluster after an Automatic Failover.

Restoring a Cluster after an Automatic Failover

To restore an Active/Passive cluster after an automatic failover, the administrator must:

1.Bring the failed node (A) back to an operational state.

At this point, this node (A) may still be configured as an Active node. This is not the required configuration.

The required configuration is:

•The recovered node (A) will become the new Passive node.

•The standalone “Active” node that is servicing requests (B) will become the new Active node.

2.Demote the recovered node (A) to be a standalone node.

3.Configure the internal interfaces of both nodes, see Prerequisites.

4.Promote the standalone “Active” node (B) to be a formal Active node.

5.Join the standalone node (A) as a Passive node to the Active node (B).

This will reset all data on the Passive node (A).

6.Monitor the synchronization of the nodes using cluster status.

Performing a Manual Failover with a Dual-Resolving DNS Entry

You can configure the DNS entry for the cluster hostname to resolve to the IP addresses of both the Active and Passive nodes. In this scenario, all PCS/PPS appliances can communicate with either functioning node. In the event of a failure, one node is still available, and the appliances will find it automatically.

You may want to perform a manual failover, because either:

•You do not have automatic failover configured, OR

•You have automatic failover configured, but you need to perform scheduled maintenance on the Active node.

To perform a manual failover using a duel-resolving DNS Entry for the cluster hostname:

1.Log into the both the Active and Passive nodes, and ensure that their services are in sync:

cluster status

2.From the CLI of the Passive node, run the following command.

cluster demote

After demoting an appliance (Active node or Passive node), reboot the appliance. The Internal port is automatically disabled. If the appliance needs to be clustered again, configure the internal interface prior to executing any clustering commands.

3.From the CLI of the Active node, run the following command.

cluster demote

4.Take the Active node out of service:

services stop

After a short time, the PCS/PPS appliances will automatically communicate with the remaining standalone node, using the dual-resolving DNS.

5.Proceed to log in to the Pulse One admin UI by accessing the cluster hostname.

To revert to using the original node as Active, you must bring the original Active node back into service and perform another manual failover before manually reforming the cluster.

Performing a Manual Failover Using a Single DNS Entry

This procedure may be required if the DNS entry for the cluster hostname resolves to a single IP address. That is, the IP address for the Active node.

Where a dual-resolving DNS entry is used for the cluster hostname, see Performing a Manual Failover with a Dual-Resolving DNS Entry.

You may want to perform a manual failover, to either:

•Recover from the failure of an Active node, OR

•Perform scheduled maintenance on the Active node.

To perform a manual failover using a single DNS Entry for the cluster hostname:

1.Log into the both the Active and Passive nodes, and ensure that their services are in sync:

cluster status

2.From the CLI of the Passive node, run the following command.

cluster demote

3.Navigate to the DNS server. Make the Pulse One FQDN hostname resolve to the IP address of the Passive node EXT Interface. That is, the IP in the DNS entry should now be for the Passive node's EXT interface.

4.Next, set Pulse One DNS record to the IP address (external interface) of the Passive node (which is now demoted to a standalone node).

5.Proceed to log in to the Pulse One admin UI by accessing the Fully Qualified Domain Name on the appliance.

To revert to using the original node as Active, you must repeat this procedure and then manually reform the cluster.

Performing a Manual Failover Using Interface IP Swap

This procedure may be required where your cluster is configured using external interface IP addresses only. That is, there is no DNS entry for the cluster hostname.

You may want to perform a manual failover, to either:

•Recover from the failure of an active node, OR

•Perform scheduled maintenance on the Active node.

To perform a manual failover by manually swapping the external interface IP address for both clustered nodes:

1.Log into the both the Active and Passive nodes, and ensure that their services are in sync:

cluster status

2.From the Passive node, type the following command:

cluster demote

3.Run the following command to change the IP of the Passive node (EXT interface) to the IP of the Active node:

network interface external --ip <EXT IP of ACTIVE node>

--netmask <netmask of ACTIVE node> --gateway <gateway IP address>

4.Proceed to log in to the Pulse One admin UI by accessing the Fully Qualified Domain Name on the appliance.

To use NFS for log storage, both Active and Passive nodes must be configured to use separate NFS mounts, see Configuring Pulse One as an NFS Server.

To revert to using the original node as Active, you must repeat this procedure and then manually reform the cluster.