Difference between batches of non-matching and registered Cassandra in negative cases

Question

I understand the main difference between the LOGGED and UNLOGGED parties in Kassandra in terms of atomicity. In essence, LOGGED parties are atomic, while UNLOGGED are not. This means that all instructions in the LOGGED batch are executed (or not executed) together.

In the case of the UNLOGGED batch, if something goes wrong during the write operation of the layout statement, I know that the statements that are already executed are NOT rollbacks, but does Cassandra notify the entire lot of the driver to refuse?

Answer 1

Thus, registered batches use a journal to record a batch operation and then perform it by deleting it from the journal when it is successful. Unlogged is still a batch operation, but at no additional cost in the log. In small quantities, logging is fine, but as the scaling increases, this batch log can grow and become a problem. Datastax docs actually cover packaging and some examples:

https://docs.datastax.com/en/dse/6.0/cql/cql/cql_using/useBatch.html

Good parties example

Bad Party Example

Generally speaking, packages are used, but I have seen that they cause performance problems when used excessively due to the fine that you pay for grouping them on the coordinator node. I often point people to this well-known blog , which also contains some useful party information.