Skip to content

Implicitly handle flushing operations

Sami Hiltunen requested to merge smh-flush-after-apply into master

The logged operations currently include flushes, and the logged flushes are performed when applying a transaction's operations. While this is nice in the sense that all operations are logged, it's not entirely necessary to log the flushes as the node applying the operation can determine from the operations which paths need flushing. When WAL archival comes around, we might be loading large amounts of log entries from the archive. Applying them is more efficient if we don't flush in between. In this case, logging the flushes is pointless.

This MR adds logic in applyOperations to automatically flush all dirtied paths.

Merge request reports