I am using akka persistence framework recently to solve an aggregate computing problem, the amount of data is a bit large, the upstream is kafka, the downstream is kinesis, but every time when I redeploy the system, from the log and database data, every time Some data will be lost. Is there any good way to avoid this?