StephanEwen commented on pull request #8751: URL: https://github.com/apache/flink/pull/8751#issuecomment-804200670
@carp84 In FLIP-158, the checkpoint interval doesn't influence when RocksDB takes its checkpoint. Instead, RocksDB backend decides itself when to checkpoint, and can do this for example always when we estimate that a full write buffer has been accumulated (like when gathered as many bytes 0.95 x write_buffer_size). That way we avoid small SST files. What may remain are small metadata files (like few kilobytes size), but we could generally inline the metadata into the state handles directly. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
