Open Acent0r opened 3 years ago
it's not the main function of netmaker, you need to deploy the server with k3s or k8s (it does what you what) , in this way you can keep you server continue online, that is your choice, the fallback server in netmaker (if it exists) don't do this in my option, its main uses for UDP punching. so you can search what is fallback server in nat network ,this will help you to understand.
Added to triage. Not certain when we would implement this, may have to come further down in the roadmap.
Issue summary: Loss of connectivity or crash on main Netmaker server can cause loss of connection between clients, especially if using UDP punching or IP changes on some client nodes
Suggestion: Allow some client nodes also be fallback servers. Fallback servers should be static (i.e. no UDP Hole Punching). A list of fallback servers (ordered?) must be held by all the nodes on synchronization.
Fallback server acts like a normal client, except creating a local backup of db when syncing with main server.
If connectivity to the main server is lost - a fallback server takes over and marks itself as the acting server. Other nodes will find the main server is down and go down the list until they find a working fallback.
If connectivity to the main server is restored - it should synchronize the data from the fallback server, and set itself as the active server. The fallback server would become inactive again and follow normal routine.
Further thoughts: Consider other methods of fallback server behaviour. This should improve the High Availability perspective of the product overall, and maintain the (partial) connectivity in critical network failures.