There are 3 different types of detection:
- multicast. Therefore, we scream over the network and try to find other participants.
- tcp / ip: we need to have some well-known members. If one or more of these known members are online, other members may form a cluster.
- aws: we just go to aws, read out all instances in a given region, apply some filtering, and what we remain are well-known members. From now on, we rely on tcp / ip based clustering.
So, this is a brief summary of how automatic detection works.
Detection of node failure is performed based on heart rate. Therefore, every x seconds we send a message to all members of the clusters, those that cannot respond are ultimately declared dead.
source share