Potential Issue re delete post-compaction
martinsumner opened this issue · 1 comments
martinsumner commented
Placeholder for now. Investigation underway after pauses were seen on riak nodes (e.g. all vnodes on a node concurrently report a long running transaction in leveled, of the same length e.g. 8.5s), that appear to occur in the 30s following the file deletions which follow Journal Compaction.
Issue so far seen only on AWS/EBS/BTRFS, but may not be platform related.
martinsumner commented
Issue was triggered by file:delete, but was specific to the use of BTRFS, switching test to (the recommended) XFS returned stable/un-paused performance.