ClickStack's June–July release bundles two months of updates. Trace waterfall gets per-service colors, a minimap for navigating long traces, and OpenTelemetry span link support in the detail panel. Metrics gain quantile support for exponential histograms, a leaner default schema using hashed attributes in the sort key, and the ability to connect external Prometheus-compatible endpoints. Dashboard filters can now cascade (faceted search style), event patterns are available as dashboard tiles, and kiosk mode locks dashboards for wall displays. Filter autocomplete now uses ClickHouse text indexes via mergeTreeTextIndex() (requires ClickHouse 26.3+) before falling back to materialized view rollups. Alerts support consecutive-window evaluation to suppress transient spikes, reporting a PENDING state while accumulating windows. The Datadog receiver for the OpenTelemetry Collector lets existing Datadog Agents send data to ClickStack without re-instrumentation. The MCP server gains metric discovery tools and source/webhook management. Smaller additions include categorical bar charts, pinned chart tooltips, builder-to-SQL conversion, table column coloring, alert annotations on tiles, and Browser RUM dashboard enhancements.
Table of contents
New contributors #Trace waterfall improvements #Connect to an external Prometheus datastore #Faster filter values and autocomplete #Cascading dashboard filters #Consecutive-window alerts #Exponential histogram metrics #Categorical bar charts #Event patterns as a dashboard tile #Read-only kiosk mode #Saved searches and webhooks over the External API #A leaner and faster metrics schema #Datadog receiver for the OpenTelemetry Collector #MCP server improvements #Little but useful things #Conclusion #Questions this post answers
What ClickHouse version is required to use mergeTreeTextIndex() for filter autocomplete in ClickStack?
The mergeTreeTextIndex() path requires ClickHouse 26.3 or later. On earlier versions, ClickStack's filter router skips it and falls back to metadata rollup materialized views instead, so filters continue to work. The function reads terms already stored in the index without scanning underlying data parts, and works with text indexes on maps as well as native columns like TraceId. Teams upgrading ClickHouse deployments track version-gated features like this on daily.dev.
How does ClickStack calculate quantiles for OpenTelemetry exponential histogram metrics?
ClickStack normalizes all bucket counts to the smallest scale present across series before aggregating, combining finer buckets into coarser ones while keeping offsets aligned. It then converts cumulative counts to deltas by aligning each bucket index with the previous data point in the same series. After summing buckets across series per time bucket, it uses log-linear interpolation within the target bucket to estimate the quantile value, matching Prometheus behavior. Developers migrating Prometheus metric workloads to ClickHouse find relevant comparisons and updates on daily.dev.
How do consecutive-window alerts work in ClickStack and what is the PENDING state?
Consecutive-window alerts require a threshold to be breached across a configurable number of back-to-back evaluation windows before a notification fires. For example, requiring three consecutive one-minute windows means the alert only pages after three successive breaches, filtering out transient spikes without raising the threshold. While accumulating the required windows, the alert enters a PENDING state, distinguishing it from an alert with no active condition. The setting is numConsecutiveWindows, available in the alert editor and external API. SREs tuning alert sensitivity for on-call workflows follow ClickStack and observability tooling news on daily.dev.
6.7K Impressions1 Comment