From the HA configuration shown for HQ-NGFW-1:
set memory-based-failover enable
set memory-failover-threshold 70
set memory-failover-monitor-period 50
set memory-failover-sample-rate 10
set memory-failover-flip-timeout 60
set override disable
set priority 200
From the performance status outputs:
HQ-NGFW-1 memory used is 90% (well above the configured threshold of 70%)
HQ-NGFW-2 memory used is about 48.7% (well below the threshold)
What happens in FortiOS 7.6 with memory-based failover
When memory-based failover is enabled, FortiGate monitors memory utilization. If the unit’s memory usage stays above the configured memory-failover-threshold for the configured memory-failover-monitor-period, the cluster triggers a failover away from the unit under memory pressure.
Threshold = 70%
HQ-NGFW-1 is at 90%, so it violates the threshold.
Monitor period = 50 seconds.
The administrator observed for 55 seconds, which is longer than 50 seconds, so the condition is met for long enough to trigger failover.
The memory-failover-flip-timeout 60 is used to prevent rapid back-and-forth role changes (flapping) after a failover decision; it does not prevent the initial failover from occurring once the threshold breach persists for the monitor period.
