v0.1.17
Released at 2027-10-15
Update Note 1: vmestimator: vmestimator now refuses to start if the -config file contains streams with identical labels, group_by, interval and filter. Such streams produce identical time series, so differences in other fields (group_limit, buckets, hll_precision, hll_sparse, calculate_churn, min_cardinality) were ambiguous. The group_by order is ignored, and an omitted interval is treated as the default 5m. Before upgrading, merge or remove duplicate streams in the config.
Update Note 2: vmestimator: the churn_interval stream config option is replaced with calculate_churn: true. Churn is now always calculated over the stream interval, and the churn_interval label is removed from cardinality_churn_ratio metrics. Configs that still contain churn_interval fail to load. Before upgrading, replace churn_interval: <duration> with calculate_churn: true and update queries that rely on the churn_interval label. In cluster mode, upgrade all nodes together: snapshots from older storage nodes do not carry the churn flag, so churn is not reported for them until they are upgraded.
-
SECURITY: upgrade Go builder from Go1.26.6 to Go1.27.2. See the list of issues addressed in Go1.27.2.
-
FEATURE: dashboards: add "Trend" section with a long-term cardinality overview (90 days). Add dedicated "Global Churn %" and "Global Churn" (absolute) panels. Add "Top 10 Churn %" panel per group_by_keys. Remove the
instancevariable. -
FEATURE: vmestimator: add per-stream
min_cardinalityconfig field to override the global-cardinalityMetrics.minCardinalityflag for individual streams. When unset or0, the global flag value is used. A negative value disables the threshold for the stream. See #46. -
FEATURE: vmestimator: support reloading the
-configfile without restart onSIGHUPsignal, on request to/-/reloadendpoint (protected with-reloadAuthKey) and every-configCheckInterval. Streams with unchanged settings keep their accumulated estimates and churn state, while added or changed streams start from scratch. See #57.