Percona XtraDB Cluster MySQL operator K8s with ha-proxy for failover

Hi Team,
I have a use case where i m having two percona MySQL operators in the same gke cluster and i have a centralised ha-proxy, what i m trying over here is i want to configure these two percona mysql operators in the centralised ha-proxy and in case if one operator goes down i want to redirect request to secondary operator.

  1. I was able to add both the operator in ha-proxy config and ha-proxy redirected request only to primary and once primary goes down then the request where getting redirected to secondary so far good
  2. but problem started when i brought back the primary again then all over sudden the request where started redirecting towards primary now the problem is about splitbrain since without syncing the primary to secondary writes where started on primary operator
  3. Is there anyway or configuration in ha-proxy to avoid sending request to the primary operator in-case by chance if primary comes back again ?
  4. As of now the syncing the data in manual so i don’t want the request to point towards primary without syncing back the data from secondary operator.
  5. so Is there any configuration which i can add in ha-proxy which can prevent pointing back to primary automatically ?
    I notes after i brought back primary i still saw the request where going towards secondary (since the switch over happened when i brought down primary as timeout) This was the thing i needed but when i intentionally went and restarted ha-proxy pod it again started pointing towards primary

Thanks in advance