Exploring the idea of using “append” instead of creating new files with HDFS every few minutes. Are there particular design decisions / considerations?
There’s certainly a history of append with HDFS, mainly, earlier versions of Hadoop warn strongly against using file append semantics. -Charles
