- Database - Master/Master in different DC;
- 2 Zabbix Server HA nodes;
If link between DC for some reason fail more than silence interval for the HA node, each HA node will acquire DB as a primary and create two separate instances.
- killswitch if amount of alive nodes less than defined value;
- 2 node HA cluster should fail to switchover if second node is not visible to it through database (in current implementation), but current active node should work until restarted;
- also if both nodes at start cannot see 2 of 2 they should fail to start;