What is the difference between an active-passive and an active-active cluster?
Short answer
In an active-passive cluster, one device carries the traffic, the other waits in reserve and takes over only on a failure. In an active-active cluster, the two devices carry traffic at the same time; if one fails, the other takes on its load as well. In TR7, both models are set up with the same cluster setting, because the Master/Slave role belongs not to the device but to each VIP: when the VIP is added, or when it is edited with Edit IP, the device on which the VIP will normally run is selected in the Master/slave field in the IP table (default Master, that is, the device logged in to). If all VIPs are Master on the same device, the setup is active-passive; if the VIPs are split between the two devices, both devices receive traffic and the setup is active-active. In active-active, when one device fails, all VIPs gather on the remaining device; therefore each device is planned with the capacity to carry the total load of the two devices on its own.
Where It Is Used in TR7
| Where | Field / label | What for |
|---|---|---|
| Network → Interfaces → Add IP → Ip Type = VIP | Master/slave (default Master) | Selecting on which device the VIP will normally run |
| Network → Interfaces | VIPs with the (M) suffix on the Master side and (S) on the Slave side; two columns per device (Cluster mode differences) | Seeing on which device each VIP is Master |
| Configuration Menu → HA Clustering & Sync card | Card description: active-active or active-passive HA clustering (HA Clustering & Sync) | That both models work with the cluster set up on this screen |
Comparison
| Active-passive | Active-active | |
|---|---|---|
| Carries the traffic | One device | Two devices |
| Setup in TR7 | The same device is Master on all VIPs | One device is Master on some of the VIPs, the other device on the rest |
| On failure | The backup device takes all VIPs | The remaining device takes its peer's VIPs as well |
| Capacity plan | Each device must be able to carry the whole load | Each device must be able to carry the total load of the two devices |
| Maintenance | The passive device is taken into maintenance without affecting traffic | Before maintenance, the device's VIPs are moved to the peer device |
Details of the role, the status texts on the screen and changing the role: What is the Master/Slave role on a VIP?
Common Misconceptions
- "A separate cluster mode is turned on for active-active." In TR7 the cluster setting is the same; active-active is set up by distributing the Master device of the VIPs across the two devices.
- "Active-active doubles the capacity." In the normal state the load is split between the two devices, but on a failure a single device carries the load of the two devices; the capacity is planned accordingly.
- "The device with Cluster ID 1 is the active device." Cluster ID determines only which device initiates the connection in pairing; which device carries traffic is determined by the Master/slave selection of the VIPs.
How it is done in the documentation
