How does a cluster work?

Short answer

A TR7 cluster consists of two devices, and two separate mechanisms work together. The first is configuration synchronization (HA Clustering & Sync); the devices connect over a dedicated Sync Interface, and a change made on one device is transferred automatically to the other device by default (Manual Sync Mode is also supported). The second is VIP failover; each VIP is Master on one device and Slave on the other, and the devices communicate with VRRP over the Interface IP pair selected for VIP Communication. If the advertisements from the Master stop, or if the link or gateway check fails according to the selected VIP Transition Method, the VIP moves to the peer device. Since the role belongs to the VIP, not to the device, an active-passive setup, or an active-active setup with different VIPs, can be built.

Two Mechanisms

	Configuration synchronization	VIP failover
What is moved	Load Balancer, network, firewall and management access, users, GTM and other settings	The shared VIP address
Over which path	Sync Interface (selected on each device in Hardware Management)	VIP Communication pair (selected in the Edit Interface form, on the interface where the VIP is)
Direction	Bidirectional; a change goes to the other side regardless of the device on which it is made	The VIP runs on the device where it is Master; when that device goes out of service, it moves to the peer device
Details	Is the configuration automatically synchronized between cluster members?	What is the Master/Slave role on a VIP? · When a failed device comes back, does the VIP automatically move back to it?

How Does It Work?

1. Pairing: Clustering Enabled is turned on on both devices; the device with Cluster ID 1 initiates the connection, and the one with 2 waits. While connecting, the CPU cores, memory and interface names are compared; if they differ, the connection is not established.
2. Synchronization: A change saved on one device is sent to the other device and applied there as well. The Clustering Enabled, Cluster ID, Manual Sync Mode, Sync Interface and Device DNS settings are device-specific and are not synchronized. If Manual Sync Mode is on, the changes are sent manually with Synchronize.
3. VIP roles: When a VIP is added, the role of the device logged in to is selected in the Master/Slave field; the peer device automatically takes the opposite role (What is the Master/Slave role on a VIP?).
4. Monitoring and transition: The devices exchange VRRP advertisements over the VIP Communication pair (VIP transition). If VIP Transition Method = Only VRRP (default), only the stopping of the advertisements triggers the transition; the transition time is set with the VIP Transition Duration (s) field (default 3 seconds). In the TR7 Link Check, TR7 GW Check and TR7 Link and GW Check methods, TR7 additionally checks the link of the interface and/or the default gateway of the route table; the transition decision is made according to the check results of the two devices: if the check of only one device succeeds, the VIP runs on that device; if both succeed or both fail, the VIP runs according to the configured Master/Slave role.
5. Traffic: The Load Balancer configuration exists on both devices; the traffic is received by the device on which the VIP is active at that moment. For the vService to be redundant, its Frontend IP must be a VIP (How is a VIP defined after the cluster is set up?).

Where Is It Configured and Monitored in TR7?

• Setup: Settings → Device Management → Hardware Management (Sync Interface) and HA Clustering & Sync (Clustering Enabled, Cluster ID, Manual Sync Mode) (Is TR7 itself a single point of failure; how do I set up Load Balancer redundantly?).
• Cluster Status: The Cluster Status, Sync Interface and Sync. Summary rows on the HA Clustering & Sync screen.
• VIPs: On the Network → Interfaces screen, the two devices are shown in separate columns; the VIPs are listed with the (M) suffix on the Master side and the (S) suffix on the Slave side (cluster mode differences).
• vService list: The Status circle is split into two halves; the left half shows the device logged in to, and the right half shows the peer device (What does a status circle split in two mean in the vService list in a cluster?).

Common Misconceptions

• "The device with Cluster ID 1 is the Master." The Cluster ID only determines who initiates the connection during pairing; Master/Slave is selected separately for each VIP.
• "The Sync Interface and VIP Communication are the same thing." The Sync Interface is for configuration synchronization; VIP Communication is the Interface IP pair on the interface where the VIP is, and it is for VRRP.
• "Once the cluster is set up, the vServices become redundant on their own." The Interface IP is specific to each device and does not move to the peer device; the Frontend IP of the vService must be a VIP.
• "Any number of devices can be added to a cluster." A cluster consists of two devices; the Cluster ID can only be 1 or 2.

How it is done in the documentation

• HA Clustering & Sync
• Interfaces · The VIP Transition Mechanisms
• Edit Interface · VIP Transition Method

Related questions

• Is TR7 itself a single point of failure; how do I set up Load Balancer redundantly?
• How is a VIP defined after the cluster is set up?
• What is the Master/Slave role on a VIP?