Skip to main content

v0.6.0: open at both ends

avuru obs team
Maintainers

v0.6.0 is about the two ends of the pipe. Getting data in no longer means re-pointing every sender you own, and getting a second cluster on the screen no longer means a second instance. Adopting avuru obs became a reversible decision instead of a migration.

  • Keep the senders you already run. Four more push protocols land beside OTLP: Jaeger (gRPC and thrift/HTTP), Zipkin, Prometheus remote-write and Loki push — one values flag each, all off by default, so an install that wants none renders exactly as before. None of them is a side door: every enabled receiver joins the same tenant stage as OTLP, so per-project ingest keys are enforced identically whatever the wire protocol.

  • Leave the same way you came in. gateway.forward.otlp and gateway.forward.kafka dual-write everything the gateway ingests to a second destination — the backend you run today, or a topic another team owns. The forwarders always carry a bounded sending queue, so a dead second target cannot backpressure the write path to storage, and Kafka SASL credentials come only from an existing Secret.

  • The compatibility claim is a gate, not a sentence. Every protocol above is exercised in CI: a real Jaeger batch, a Zipkin span, a snappy-framed remote-write sample and a Loki push go through the receivers a real Helm install renders, then the rows are asserted in storage and the forwarded trace is checked arriving at a stand-in legacy backend.

  • One project, several clusters. Member projects make a project read the union of the projects it aggregates — services, map, traces, logs, metrics, errors, profiles, energy, alert history. Nothing about the members changes: they stay separately queryable and separately granted, and a viewer still sees only the members they already had. The install now matches: every component has a switch (hub.enabled, ui.enabled, gateway.enabled), so a secondary cluster installs the ingest half alone against the central store, and combinations that cannot work are refused at install time with a sentence saying why.

  • A project can keep less than the install does. Give any UI-managed project a shorter retention window and a background sweep trims that tenant hourly, scoped by project. A window longer than the install-wide one is refused rather than quietly ignored — the shared table TTL would drop those rows first whatever the project asked for.

  • What one project holds, not just what the install holds. Settings → Storage now answers per project: rows, an estimated size, the ingest rate over the last hour, how far back the data goes, and the retention that actually applies. Size is the one estimate, and the column says so — storage parts hold every project's rows together, so a project's share can only be apportioned by row count.

Sustainability​

  • The energy numbers now say which nodes they came from. /green counted its nodes — known, measured, estimated, absent — but never said which node spent what. The coverage panel gained a per-node table with the measured/estimated split kept visible per row, and nodes reporting nothing listed at 0 Wh rather than omitted. Carbon budgets now also resolve their own deliverability (alerting off, no channel, unknown channel) instead of rendering identically whether or not they can reach anyone, and a budget aimed at a service group nothing rolls up to says so instead of evaluating forever at zero. A RAPL validation runbook now covers checking every hop on real hardware.

Fixed​

  • The hub reported the retention it was built with, not the one you configured. retention.* reached the migration job but never the hub itself, so an install keeping 30 days of traces had a hub still answering with the 7-day default — which Settings → Storage read as TTL drift and reported as a migration to re-run. Both sides now render from one chart helper, and a template assertion fails if either loses it.

Upgrading from v0.5.x is a normal helm upgrade; schema migration 0019 applies automatically (hub.autoMigrate), and every new receiver, forwarder and component switch is off or unchanged by default. See the Releases page and the GitHub release.