if you go to 15 you will have 16 copies of the data and it does not make sense. I think that you may have the concept confuse.
The ideal K-safe = 1. You keep one copy of the data in a 'adjacent' node so if you lose a node in the cluster if the adjacent node is UP, the database still up and that adjacent node answer the questions that the node down should have. In that scenario you can loose 15 nodes, ONLY if the adjacent nodes still up.
If you have k-Safe = 2, you keep 2 copy of data but the number of nodes that can go down still the same, 15 , the only difference it is which one.
Does make sense, increasing the k-safe does not mean that you increase your changes to keep the database up, you just change which nodes become critical once that nodes go down. But at the cost of loading more data.
To see what nodes are critical, once that you have a node down you can query select * from critical_nodes; and that will give you the list of nodes that if one goes down the database will shutdown because loss of k-safe.
I think that there is not a limitation on the number of nodes. You need to ensure that the network connections between them meet the requirements ( use vnetperf for that) and it should work. If you have network latency it can affect query performance and if your UDP traffic is not stable, you could get some nodes dropping the cluster. HTH Eugenia
Comments
The ideal K-safe = 1. You keep one copy of the data in a 'adjacent' node so if you lose a node in the cluster if the adjacent node is UP, the database still up and that adjacent node answer the questions that the node down should have. In that scenario you can loose 15 nodes, ONLY if the adjacent nodes still up.
If you have k-Safe = 2, you keep 2 copy of data but the number of nodes that can go down still the same, 15 , the only difference it is which one.
Does make sense, increasing the k-safe does not mean that you increase your changes to keep the database up, you just change which nodes become critical once that nodes go down. But at the cost of loading more data.
To see what nodes are critical, once that you have a node down you can query select * from critical_nodes; and that will give you the list of nodes that if one goes down the database will shutdown because loss of k-safe.
Hope this helps to understand the concept.
Eugenia
What is the maximum number of nodes in a cluster on cloud?