CDC historically benchmarked changefeed behavior in roachtests by testing against both the kv0 (write-only) and kv100 (read-only) workloads. kv100 is useful to monitor changefeed performance when no data is processed to make sure there are no regressions.
Due to test flakiness, kv100 was disabled in the benchmarks. We should consider reenabling it at some point, since it's still a somewhat useful data point.
blathers-crl[bot]
commented
12 hours ago
CDC historically benchmarked changefeed behavior in roachtests by testing against both the kv0 (write-only) and kv100 (read-only) workloads. kv100 is useful to monitor changefeed performance when no data is processed to make sure there are no regressions.
Due to test flakiness, kv100 was disabled in the benchmarks. We should consider reenabling it at some point, since it's still a somewhat useful data point.
Jira issue: CRDB-44800