If a node binds the two network ports of one NIC, it supports automatic failover of faulty links—when a network port fails, other ports can automatically take over the failed port to forward traffic.
When using the active-backup bonding mode, the switchover of a failed link completes within 4 seconds. The cluster virtual machine will then experience a brief delay, with a slight decline in I/O performance, but it will return to normal after the switchover. Conversely, in non active-backup bonding modes, the abnormality of a single network port usually results in the node's maximum write bandwidth being halved.
If all the bound network ports fail simultaneously, the node will be unable to provide connectivity for the corresponding network plane (management, storage, or migration).
