what different effect between "changelog.enabled" and "cdc.enabled"?

[li-ang-666]

as the title, for COW table, which one should I use?

[danny0405]

• changelog.enabled implies changelog mode, which requires the upstream source is a changelog stream; it is only enabed for MOR table, it has the best performance for changelog stream ingestion and consumption(Hudi will store the operation for each record);
• cdc.enabled implies cdc mode, it does not require upstream source to be a changelog stream, it is enabled for both COW and MOR table, Hudi will generates the changelog on commits(for COW) and compaction (for MOR).

[li-ang-666]

ok! I knew, and how could I only read the changelog data, without any snapshot data? I Read COW table and engine is Flink, what's different betwen 'read.type = incremental' and 'read.streaming.enabled = true' ?

[danny0405]

if your upstream data source is a changelog(for e.g, binlog from MySQL), changelog mode is suggested, and the output for streaming read of the Hudi table would be always a changelog. Otherwise use cdc mode, when cdc.enabled is true, the streaming read will switch to read the changelogs(generated by Hudi).