v1.4.0 ships native OTLP/HTTP ingestion, Kubernetes log and metric collection templates, alert silencing and grouping, plus several tracing and search fixes.
v1.4.0 is our biggest release this quarter, moving OpenTelemetry from "compatible" to "native" and filling in Kubernetes collection and the alerting experience. Here are the full notes and upgrade guidance — read the last section before you upgrade.
This release speaks OTLP/HTTP natively, so any SDK or Collector that emits OTLP data connects directly, with no intermediate adapter to configure or keep in sync. All three signal types from OpenTelemetry 1.x SDKs — traces, metrics and logs — are supported, so an existing OTel deployment can be re-pointed at Observe by changing a single endpoint. Teams already instrumented can start reporting immediately; teams that aren't can drop in our Java/Go auto-instrumentation agents, which turn HTTP, RPC and database calls into traces automatically, with no code changes. The Java agent currently covers Spring Boot, Dubbo, gRPC and mainstream JDBC drivers; the Go agent covers net/http, gRPC and common ORMs. If you've been putting off tracing because instrumenting every service looked like a project, this is the release that removes the excuse.
Kubernetes collection also lands here: built-in templates let you deploy a DaemonSet from the console with one click, capturing container logs and cAdvisor metrics automatically. Logs arrive with pod, namespace and label metadata attached, so you can filter by namespace right away. Helm installation is supported, and both containerd and Docker runtimes are covered — saving you the time of hand-writing Collector configuration from scratch.
Alert rules now support silencing windows, so scheduled maintenance stops waking people up, and grouping by service or alert type, so a single incident no longer floods you with hundreds of near-identical notifications. A new alert history page shows each alert's trigger and recovery timeline, which makes postmortems faster and records how long an incident actually lasted. Rules also gain a for duration field, which filters out transient spikes and cuts the false positives that train people to ignore alerts in the first place.
LIKE wildcard escaping bug in SQL-like searchThis release changes the storage-layer index structure. After upgrading, indexes rebuild automatically in the background, and search may be briefly slower during that window — run the upgrade during off-peak hours. Back up the database before upgrading, and after the upgrade confirm that all alert rules were restored. Custom Collector configurations from v1.3 remain compatible but should be migrated to the new built-in templates; the old templates will be deprecated in v1.5. Known issue: hydration of extremely large traces (over 5,000 spans) can still lag, to be addressed in v1.4.1. If you hit any issue during the upgrade, use the Help button in the console's top-right corner — include your version number and error logs to speed things up.
This release merges over 40 pull requests; thanks to everyone who joined the beta. v1.4.1 will focus on the large-trace hydration lag and add more aggregation operators to alert rules.