Zero Downtime Mastering Cassandra Live Migrations at Scale
summary
Migrates over 1+ billion records from Cassandra to a DataStax serverless target while preserving zero downtime by using a proxy (open‑source zdm) for dual writes and a customized Spark migrator built on the Cassandra Spark connector. Builds a separate Spark cluster to split token ranges, tunes migration rate, batch size and splits per table, runs 10% production tests, runs row/column validation jobs (with autocorrect option), and maintains HA proxy, observability and rollback until cutover succeeds.
AI-written summary · about
chapters (10)
- 0:00Introduction to migration
- 1:49Migration guiding principles
- 3:53Handling live traffic and data
- 6:45Choosing migration tools
- 10:16Pre-migration preparations
- 15:13Migration process overview
- 16:57Critical challenges in migration
- 21:49Migration phase considerations
- 22:53Validation and optimization
- 23:14Final cut-off and conclusion
related talks
31mZero Downtime Migrations at Scale
32mMigrating Applications Between Cassandra Clusters with No Downtime
24mMigrations under Production Load: How to Switch...
34mLarge Scale NoSQL DB Migration Under Fire
35mCassandra Data Migration with Dual Write Proxy
35mReplacing Cassandra: A Digital Transformation for the World’s Largest Digital Identity Network
37mDatabase Schema Migrations with Zero Downtime
34mZero Downtime Database Migrations and Deployments