We are serving grafana on k8s pod.
The ha configuration was referred to the above document.
As you can see in the screenshot, duplicate records remain.
how can i solve it?
I see the same thing. Did you manage to solve this?
Hi! 
Grafana, even in HA, evaluates alert rules on all replicas. What HA does is deduplicate notifications, but it does not deduplicate evaluations. That means you should expect to see each state change N times, where N is the number of replicas.
These Normal states are because the rule was edited/updated - rather than because the alert was below the threshold. You’ll see Normal(Updated) in the case, instead of just Normal.
