Even after the refactoring of riak transfers, issues with timeouts of handoffs have still been observed. These issues have been limited to situations where the key count gets very large on an individual vnode - and the work backlog within the vnode increases causing frequent stalling.
None of the tuning options were previously exposed in riak.conf. To avoid confusion, just one is promoted here. The default has been changed to 500 (in most cases the batch size threshold will be applied before the count with this default).
Even after the refactoring of riak transfers, issues with timeouts of handoffs have still been observed. These issues have been limited to situations where the key count gets very large on an individual vnode - and the work backlog within the vnode increases causing frequent stalling.
None of the tuning options were previously exposed in riak.conf. To avoid confusion, just one is promoted here. The default has been changed to 500 (in most cases the batch size threshold will be applied before the count with this default).