If Cassandra is down, during that downtime any data wrote to KairosDB is lost. There is no retry, no backlog or error sent back to the client.

I would expected that KairosDB retry writing data to Cassandra when it's back online and/or reply an error to client trying to push data to KairosDB.

KairosDB 1.1.2-1 is using the default configuration (only kairosdb.service.datastore is changed to use Cassandra)
Cassandra 3.9 is using default configuration but I've to run "nodetool enablethrift".
I'm writing continuously (every 10 seconds) to KairosDB using the REST API and always get a 204.
To reproduce, I restart the Cassandra node and see that some value are missing.