6 Configuring for High Availability
Standalone controller operation provides management for the OpenFlow switches in a network.
However, it does not provide high availability (HA), with the result that a controller failure leaves
the network in an unmanaged state. Configuring a team of controllers and one or more
corresponding controller regions creates a high availability network with failover capability,
resulting in a continuously managed network in the event that a controller in the team goes down.
You can view your team configuration using the UI, see
“Viewing your team configuration using
High Availability best practices
•
Ensure the team and region configuration meets all of the requirements for teaming. For
details, see
“Requirements for teaming” (page 100)
.
•
Ensure that IP routing configuration in the controller domain enables the controller team IP
address to be reached from all areas of the domain.
•
If any application installed on the controller uses the Cassandra database, run the Cassandra
nodetool repair
command every 10 days. For details, see
Cassandra database nodes using
•
Before you create the team, configure NTP such that all the controllers that will be in the
team use the same local NTP server. Using the same NTP server for all controllers helps
to ensure that the controller clocks remain synchronized. Keeping the system clocks
synchronized is especially important for applications that use the Cassandra database. For
details, see
“Configuring controllers to use the same local NTP servers” (page 103)
•
Either use the same Keystone server for authentication for all three controllers in the team,
or use a local Keystone server for each controller in the team. For details on security, see
•
To use TLS connections for communications between the switch and the built-in OpenFlow
controller module of the HPE VAN SDN Controller, Hewlett Packard Enterprise recommends
that all controller and device certificates be signed by the same CA. For information about
configuring TLS on a switch, see the documentation for the switch.
•
If you are configuring web proxy server settings, ensure that you include the team IP address
and the IP addresses for all three controllers in the team in the
/etc/init/sdnc.conf
file. After editing the
sdnc.conf
file all three controllers will need to be restarted. For details,
see
“Obtaining applications from the Hewlett Packard Enterprise SDN App Store” (page 34)
About teaming for High Availability
•
Each controller belonging to a team is a
team member
. To centralize team management
and control, one controller is elected by the team as the
team leader
.
•
Teaming is configured on one controller and is automatically propagated to the other
controllers in the team, regardless of which controller becomes the team leader.
•
After a team is configured, any configuration changes will propagate to each controller.
•
If the team leader goes down, another active controller becomes the team leader. If a team
leader that failed recovers and rejoins the team, it rejoins the team as a team member and
does not resume team leadership.
•
Each controller in the team has its own IP address, which is the IP address of the machine
on which the controller is installed. In addition, the administrator configures a separate IP
address—called the
team IP address
—to represent the team as a whole. The team IP
address is active on the team leader. If the current team leader goes down, the failover
process includes keeping the team IP address active on the new team leader.
High Availability best practices
99