Round 12 Field Validation
Round 12 field validation
Section titled “Round 12 field validation”An observability-focused round on a live 3-node Cassandra 5.0 cluster.
Verdict: ALL GREEN. Every R11b baseline held or improved, the round’s headline — the #2419 in-process saturation gauges — is now visible under overload where R11b showed only zeros, and the #2436 large-cell fix produced a legitimate row-count increase. No restarts, no OOM, no regressions.
Highlights
Section titled “Highlights”- Regression parity vs R11b — held or better. Warm
LIMIT/point-read latencies sit at the ~JDBC floor, warm throughput andcount(*)wall time at parity, zero cold index parses (#2412). - #2436 large-cell fix confirmed — identical data returned +8,489 rows vs R11b: rows with a ≥~1MB single cell that were previously silently dropped now read correctly. A fix, not a discrepancy.
- #2419 saturation gauges visible — during an 80-thread scan overload the new in-process gauges produced a legible saturation trace (blocking tasks, egress channel depth, admission ratio, threads, fds, RSS), all rising under load and returning to baseline after drain. In R11b these read a flat, invisible 0.
- Fan-out & memory bounded — 8-thread load fans across all 3 flight pods; memory stays bounded (idle 3 Mi → ~603 MB heaviest pod at 80-thr overload), 0 OOMKills / 0 restarts.
Run metadata: flight
round12@sha256:2433dde8· connector 0.14.3 · Trino 481 · Cassandra 5.0 RF=3 · 3× i4i.xlarge · ~1.93M partitions/node · 2 SSTable gens · 2026-07-15.