When master node goes down...

tmdoit

What is your solution for this kind of scenario? Election (manual or auto) different master and doing reconnection for all nodes?

Bernardo

Currently LiteSync only supports a centralized star topology. If the primary node goes down, the network will go down too. The secondary nodes will reconnect after the primary node restarts.

Note that these nodes do not lose access to the local database. Read and write access continue normally. These modifications will be synchronized after the re-connection.

Support for many primary nodes may come later. Now the work is focused on other things.

tmdoit

I've asked because I'm thinking about the situation where for example two people will modify the same invoice and save it - you can't check if operation is valid (document wasn't modified before, after network goes down).
What you can do is to check before saving (in this case) that connection to master is alive and if it's not - forbid the operation.
This is the solution for this use case?

Bernardo

I don't know how are your specific checks for "validity", but in this case both modifications will be applied. The last one to arrive in the primary node can overwrite the changes from the previous update if they were made on the same field. If each node modified a different field then no data is overwritten.

Paradyse

How does it know the order the queries were done? Is it through a datetime stamp with the rowid?

Bernardo

The commands are applied on the order they arrive on the primary node. It does not matter the time in which they were executed locally if off-line.