Sylvain Lebresne joins host Cedric Lunven introduce NodeSync, a new approach to anti-entropy repair in DSE 6. They discuss the simplicity NodeSync brings to operational operations, improvements in performance for repair and how well it is integrated with DSE through CQL and OpsCenter.
Highlights!
0:15 - Cedrick welcomes Sylvain to the show. Sylvain a longtime member of the DataStax core DB team and an Apache Cassandra PMC member
1:10 - Explaining the concepts of anti-entropy repair in Apache Cassandra, including Merkle trees and streaming
3:39 - Limitations of anti-entropy repair include performance impacts, over-streaming of data and the complexity of managing repairs
6:11 - NodeSync is a replacement in DSE for the traditional Cassandra repair process. It leverages Cassandra’s architecture to make repairs highly available
7:45 - NodeSync is enabled/disabled on a per table basis. You can enable it and watch metrics in OpsCenter to see it working
9:12 - You can control the prioritization of tables in NodeSync by setting the target deadline for completing repairs per table
10:50 - Testing a feature like NodeSync includes testing on large clusters, simulating data loss, and stress testing
12:23 - Wrapping up