Doing an fsync per-record when receiving the complete HDB is a performance disaster. Among other things, if the HDB is very large, then one slave receving a full HDB can cause other slaves to timeout and, if HDB write activity is high enough to cause iprop log truncation, then also need full syncs, which leads to a cycle of full syncs for all slaves until HDB write activity drops. Allowing the iprop log to be larger helps, but improving receive_everything() performance helps even more.
31 KiB
31 KiB