Field notes from operating Cassandra in production.
October 26, 2026 · RapidMindLabs
An ALTER TABLE is thirty characters of CQL and one of the few operations that can split a running cluster's view of itself. How schema propagation works today, the DDL that is safe and the DDL that is not, and how Transactional Cluster Metadata in the 5.x line removes the class of problem.
Read moreOctober 19, 2026 · RapidMindLabs
Most Cassandra write-path incidents we audit start with a BATCH that was never supposed to be one. What logged, unlogged, and counter batches actually cost, the thresholds worth alerting on, and the rewrites that fix them.
Read moreOctober 12, 2026
Most sizing conversations start with a storage number and end with a surprise. Here is the arithmetic we actually use — replicated bytes, compaction headroom, per-node density limits, and the streaming time that decides how big a node is allowed to get.
Read moreOctober 5, 2026
Turning on authentication, encryption, and audit logging is easy on a laptop and risky on a live cluster. The phase order we use, the system_auth detail that locks teams out, and why 'optional: true' is the whole trick for TLS.
Read moreSeptember 28, 2026 · RapidMindLabs
Most Cassandra dashboards show hundreds of metrics and warn you about none of them in time. Here are the signals that lead incidents by hours or days, the thresholds we use, and how virtual tables let you answer questions at the cqlsh prompt during a page.
Read more