Open source observability, talked through with its builders.
OpenObservability Talks brings host Dotan Horvitz together with people close to open source observability projects, from the creator of Prometheus to the CNCF's CTO. Conversations range across managing OpenTelemetry collector fleets with OpAMP, eBPF-based instrumentation, telemetry for GenAI, and news from OTel Unplugged and PromCon.
PRESS PLAY
OpenObservability Talks is an interview show about open source observability. It is hosted by Dotan Horvitz and built on conversations with people who have a hand in the projects themselves. Each episode promises anything DevOps, observability and open source, and ends with the host's habitual sign-off: 'may the open source be with you.' The talk in between is technical, enthusiastic and friendly. OpenTelemetry is the thread that keeps coming back. The OpAMP conversation starts from a scaling problem: once organisations run many collectors across servers, containers and edge environments, how do they configure and update the whole fleet consistently and securely? The answer gets concrete. It covers how a supervisor process points at the collector binary, why the supervisor is usually deployed separately when standard package managers are used, and how a Kubernetes operator deploys collectors from CRDs instead. The eBPF conversation is just as specific. The host has followed eBPF for years as a route to truly automatic instrumentation. OpenTelemetry eBPF Instrumentation, or OBI, is based on a project donated by Grafana Labs, and the guest is frank about where it stands. RED metrics are generally available, tracing is still more limited, and a Linux kernel feature once used for tracing was later disabled as insecure. GenAI gets its own episode, built around the OpenLLMetry project and the point that the context carried with each call makes this telemetry far larger than traditional observability. Community events give several episodes their starting point. One was recorded just after FOSDEM and covers the first OTel Unplugged unconference held in Europe. That event took place in Brussels and brought together maintainers, contributors, governance committee members and users. A guest in that episode passes on an end user's view that OpenTelemetry is the best thing ever to happen in the observability industry, largely because data can go to different backends without changing instrumentation, while also admitting to paper cuts such as collector stability. Another episode was streamed in PromCon month, with Prometheus creator Julius Volz invited to take listeners through the long-awaited 3.0 release and wider ecosystem news. For the CNCF's tenth anniversary, the foundation's CTO and co-founder, Chris, sits down for a fireside chat about what has been achieved and where things are heading. One of Chris's favourite parts of KubeCon is hearing companies explain what went wrong and how they fixed it, such as a rebuild on OpenTelemetry that took a team from twelve engineers to four. Video is part of the format too: back from KubeCon North America in Atlanta, the host shows off the conference shirt to viewers. Some episodes stream live on YouTube and Twitch, with comments and questions welcome in the chat. Show notes carry references and guest contacts, and follow-up discussion happens on Bluesky, LinkedIn, X and the CNCF Slack. A practitioner running collectors at scale could start with the OpAMP episode, while the CNCF fireside chat gives the wider picture.
Made for: Suits a listener who runs or evaluates open source observability tools such as OpenTelemetry collectors and Prometheus, and wants release news and design reasoning from people close to the projects. It also fits someone following KubeCon, FOSDEM and PromCon who likes getting the takeaways as a conversation.
What sets it apart: Several episodes are tied to a community moment, such as the first OTel Unplugged in Europe, PromCon month or the CNCF's tenth anniversary. Because the guests have a hand in the projects, the talk runs from governance and release news all the way down to how a collector supervisor gets packaged.
On OpenObservability Talks we discuss harnessing the power of open source to advance observability initiatives for developers, DevOps and SRE practitioners around the world.We live-stream the episodes on Twitch and YouTube Live - tune in to see us live, and chime in with your comments and questions on the live chat.https://www.youtube.com/@openobservabilitytalks You can find us on X (Twitter) @openobserv and BlueSky @openobservability.bsky.social
Dotan Horvitz asks where collector supervisors give way to Kubernetes operators; your deployment plan gets more specific.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
A practical tour of OpAMP as a control layer for large OpenTelemetry collector fleets.
Operating OpenTelemetry at Scale with OpAMP keeps its focus on the practical control problem that shows up once OpenTelemetry collectors are running across servers, containers, edge environments, and similar estates.
A Brussels unconference turns OpenTelemetry's portability promise into a practical look at collector stability and backend choice.
OpenTelemetry's portability promise gets a grounded Brussels check after FOSDEM, where the conversation ties the project's open governance to a practical operational question: how teams can send telemetry to different backends without rewriting instrumentation.
A warm CNCF decade review anchored by OTEL scale, KubeCon community work, and the people who keep cloud native knowledge moving.
OpenObservability Talks uses CNCF's 10-year mark as the frame for a fireside conversation between Dotan Horovitz and Chris Sanacek, the foundation's CTO and co-founder.
A practical orientation to GenAI observability, from semantic conventions to the instrumentation gaps OpenLLMetry is trying to close.
Generative AI moves observability beyond the familiar job of watching systems and services. The harder question is how telemetry should represent the much larger context that travels through an LLM workflow.
A grounded look at OpenTelemetry eBPF instrumentation, OBI, and the practical limits around tracing.
OpenTelemetry's eBPF instrumentation gets a clear, focused technical pass in this conversation between host Dotan Horvath and Mario Mathias, principal software engineer at Grafana Labs and co-creator of the Bela project and OBI.
Power Platform news with practical static.
OpenObservability Talks uses this episode to explain OpAMP, the OpenTelemetry agent-management protocol for fleets of collectors that span servers, containers, edge devices, and Kubernetes.
["OpenObservability Talks S6E08 covers OpenTelemetry's state after the first OpenTelemetry Unplugged Europe in Brussels, held the day after FOSDEM.", "Dottan Horvitz interviews Zhuhasi, a member of OpenTelemetry's governance committee, about stability, graduation, and what the CNCF maturity model would recognize.", "Zhuhasi says OpenTelemetry now spans 11 languages, the Go SDK is about 80 percent to 88 percent better by his measurements, and the collector's two-week release cadence can leave users 20 versions behind.", "The conversation also hits Blueprints, backend lock-in, bad telemetry, Prometheus interoperability, the OTel injector, and Weaver as practical responses to end-user pain."]
["Chris Sanacek says CNCF has passed 300,000 contributors across more than 230 projects after starting with about 20 members and one project, Kubernetes.", "Dotan Horvitz uses the OpenObservability Talks 10th-anniversary fireside chat to ask how cloud native has expanded beyond orchestration into observability, platform engineering, and AI workloads.", "Chris argues that Kubernetes now acts like the operating system for cloud native, with DRA, GPUs, TPUs, FPGAs, and WASM all in scope.", "The episode closes on 2026 predictions about security-observability convergence, rising FinOps pressure on AI bills, and AI agents becoming major open-source contributors."]
["Open Observability Talks S6E06 centers on Nir Gazit, creator of OpenLLMetry and co-founder of Traceloop, as he explains why GenAI tracing needs prompts, completions, tool calls, and session context.", "Totan Horvitz opens from KubeCon North America in Atlanta, marks the CNCF's 10-year anniversary, and promises a breaking-news segment plus a personal blog post in the show notes.", "Nir separates AI observability into classic signals like latency, cost, and token counts and quality checks for hallucinations, jailbreak attempts, and LLM-as-a-judge scoring.", "He says OpenLLMetry now covers almost 40 providers, vector databases, and frameworks, while OpenTelemetry still struggles with large prompts and 8-megabyte vision payloads."]
OpenObservability Talks S6E05 centers on OpenTelemetry eBPF instrumentation, also called OBI, and Grafana Labs' Bela project.
Dotan Horvitz interviews Jeroen Snider, a creator of KEDA and Dapr, about the projects' cloud native origins and observability features.
ClickStack combines HyperDX, ClickHouse, and OpenTelemetry to help engineers connect a user session to a failing API call and its backend traces.
OpenObservability Talks host Dotan Horvitz and Chronosphere's Alok Bayde describe the CNCF's first Open Observability Summit as a vendor-neutral forum while Monitorama is paused.
Host Otan Horvitz interviews Prometheus creator Julius Volz about the Prometheus 3.0 release and ecosystem updates at PromCon.
Kelsey Hightower reflects on 2023 and discusses future trends in Kubernetes and cloud-native observability.
Host Dutton Horvitz interviews Fluent Bit maintainer Eduardo Silva about the new Fluent Bit version 2 release and updates in open source observability.
Live from KubeCon Europe, host Dutton Horvitz interviews CNCF's Head of Ecosystem Taylor Dolezal about record-breaking attendance and hot updates in cloud-native observability.
Jonah Cowell introduces a discussion with Eli Fisher about the OpenSearch project, its evolution, and upcoming features.