I was trying to understand how NodeSync does incremental repair. Does it run a sub-range repair at certain intervals(based on rate and target)?
@rajib76, please see this documentation to see if addresses your question https://docs.datastax.com/en/dse/6.8/dse-admin/datastax_enterprise/config/aboutNodesync_2.html and if not please do post an update here.
NodeSync uses a strategy similar to sub-range repairs but it doesn't do expensive merkle tree calculations or streaming between nodes. Instead it splits the token ranges a replica owns into small segments with an average of 200MB (actual size varies depending on the partitions contained in a segment).
NodeSync builds on the same synchronisation mechanism as read-repairs to determine if replicas are out of sync and read-repairing partitions.
As you already alluded to, NodeSync validations (process that synchronises table segments) are scheduled based on the target deadline and maximum rate (bytes/sec).
If you're interested, this blog post is a good overview of this DSE-only feature. Cheers!
8 People are following this question.