NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
Go modules
Go modules, #1789 by repository stars
Latest stable
v1.2.1
Last release 3 months ago
02 Aug 2026
Ships unpredictably
gaps range from 8 days to 7 months
Some releases are documented
notes for 15 of 59 stable releases
Nothing withdrawn
no release was ever pulled
5 years old
151 releases · first in 2022
Drag across the ruler to compare, or tap a release to open it.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Nothing published for this version
Nothing published for this version
…and Pion dependencies and integrating upstream security fixes, hardening STUNner's Config Discovery Service, optimizing the operator to handle large d…
We are proud to present STUNner v1.2.0, the next release of the STUNner Kubernetes media gateway for WebRTC brought to you by l7mp.io. This release includes many fixes and features to improve the STUNner experience for our production users: security hardening, observability, and improved robustness and stability at scale.
Major changes are upgrading STUNner to the latest Kubernetes, Gateway API, OpenTelemetry, and Pion dependencies and integrating upstream security fixes, hardening STUNner's Config Discovery Service, optimizing the operator to handle large deployments (+100 Gateways and UDPRoutes), and adding an opt-in pprof endpoint to the operator for live debugging and profiling. This release also includes broad dependency refreshes across STUNner repositories, documentation cleanups and clarifications, test and benchmark improvements, and many small stability fixes throughout the dataplane and the control plane.
The most visible change in this release is the merging of the upstream TURN NONCE size optimization that we implemented in pion/turn to mitigate ongoing TURN amplification attacks. In short, the problem is that malicious users can send unauthenticated TURN requests to STUNner so that the resultant TURN error response is larger than the triggering request, which makes the TURN server a tempting reflector/amplifier in spoofed UDP attacks. Our mitigation reduces the attack surface by making NONCE generation more compact, while preserving the security properties the TURN handshake needs.
If you want the full background, out earlier post is still the best deep dive: Mitigating TURN Amplification Attacks.
The Gateway Operator now exposes an opt-in pprof endpoint for live debugging and live profiling. This is disabled by default; enable by setting the pprof HTTP endpoint with the --pprof-bind-address=<address:<port> operator CLI flag and use the standard Golang profiler tool go tool pprof via a kubectl port-forward to attach to the running operator. Do not expose the profiling endpoint publicly. Profiles can contain sensitive runtime details; pprof is a debugging interface, not a service API.
The most important control-plane change in this release is the operator updater optimization. In previous versions, the updater could end up rewriting objects in the Kubernetes API server even when nothing semantically changed. This churn causes excess Kubernetes API traffic and slows convergence under load, to the point that eventually the Kubernetes API server starts to rate-limit the operator, leading to the control-plane becoming unresponsive.
In v1.2.0 we reworked the updater around resource-specific "lenses" that compare and apply only the parts of an object STUNner actually cares about. This has a very practical effect: the operator stops uploading objects to the API server that are already in the desired state. This alone made the operator about 100x faster at scale.
STUNner supports graceful shutdown: when a pod receives SIGTERM, it stops advertising readiness but keeps serving existing allocations until after active connections have drained. Unfortunately, some Kubernetes environments have a curious bug in the UDP data path: once the pod turns unready, Kubernetes may stop forwarding UDP packets to the pod too early, even though the process is still alive and intentionally draining existing sessions. For TURN/TCP this is usually survivable. For TURN/UDP it can be disastrous: sessions that should have drained cleanly are blackholed mid-flight.
STUNner v1.2.0 adds a targeted escape path for exactly this situation, which can be enabled with the --force-ready-status CLI flag. When this flag is enabled, STUNner keeps reporting ready status during the graceful termination window instead of deliberately failing readiness. Do not enable this unless you see premature TURN connection terminations during graceful shutdown, because the normal Kubernetes behavior is still the right one. But if your environment drops UDP too early during rollouts or scale-down, this flag gives you a pragmatic workaround.
On popular request, STUNner now supports JSON logging in the dataplane. We have always supported JSON logging in the operator, but the dataplane support was missing. Thanks to the recent upstream efforts of the Pion team, we now fully support JSON log format, a critical observability feature for our production users.
feature(icetester): Create GW for chosen proto only (#197) feature: Implement JSON logging backend, fixes #208 fix: Centralize STUNner logging options fix: Fix ready status reporting, fix #201 fix: Workaround Kubernetes UDP graceful shutdown bug chore: Merge the reconfigurable nonce PR from l7mp/turn doc: Change 'namespace' field requirement to optional doc: Update TURN offload doc test: Added CDS client-server benchmarks
feature: Add Prometheus metrics for renderer, updater, and operator threads feature: Add active loops metrics to detect operator hangs (#71) feature: Expose pprof endpoint for debugging feature: Filter well-known labels from propagating to K8s, fixes #70 fix: Allow to set controller-name via an env var fix: Fix route-indexer in the StaticService controller fix: Fix the parsing of the --throttle-timeout command line arg fix: Handle unresolved TLS certificate refs, fixes l7mp/stunner#215 fix: Let the service updater patch the spec instead of rewriting it fix: Prevent event loop deadlock when CDS clients are slow (#64) fix: Refactor LB service-port merging fix: Robustify Gateway API UDPRoute CRD detection, closes #69 fix: Use semantic deep-equal in node controller chore: Add config rendering pipeline benchmarks chore: Make updater logging more consistent refactor: Optimize updater by suppressing useless reads/writes refactor: Skip v1alpha2 UDPRoutes when CRD is not available (#67) test: Add bootstrap benchmark workflow for operator profiling
Enjoy STUNner, join us on Discord, and stay tuned for the detailed release notes.
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
…resources (Deployments/DaemonSets) fix: Deprecate the rbac-proxy, fixes l7mp/stunner#185 fix: Generate NoMatchingParent status when route parent is mi…
We are proud to present STUNner v1.1.0, the second GA release of the STUNner Kubernetes media gateway for WebRTC brought to you by l7mp.io. This release marks the culmination of a half-year development and stabilization cycle to bring your favorite WebRTC ingress gateway suite to a carrier-grade level.
Most of the work in this release was focused on implementing the new carrier-grade commercial features in STUNner that will allow it to scale to extremely large and performance-sensitive deployments. The open source version was not left behind though: major additions are a new ICE tester tool to simplify testing a fresh STUNner installation, new tutorials, lots of improvements to stunnerctl, plus countless improvements and fixes all around the place.
Originally, deploying STUNner required a separate installation of the stunner/stunner-gateway-operator Helm chart to deploy the control plane, plus the stunner/stunner chart that would deploy the dataplane that actually implements the TURN gateways. With the introduction of the managed dataplane mode this legacy dataplane mode has become obsolete, and it has gone mostly unmaintained in the last couple of releases. With the v1.1 release support for the legacy mode has been completely removed. In addition, the stunner/stunner Helm chart will now install the gateway operator instead of a standalone dataplane.
The open-source STUNner release has served our growing user base solidly throughout the last couple of years, making WebRTC available as a native Kubernetes service in a wide range of applications, from cloud-gaming to live streaming and conferencing. That being said, there are still a couple of issues that prevent large-scale enterprises to migrate from legacy TURN server frameworks to STUNner. Chiefly among these is performance (what else?): while Go, the programming language we used to develop STUNner, has provided solid performance for small to medium-scale deployments, the runtime overhead of the memory-managed Go runtime has proved to be a bottleneck in some extremely large use cases. In addition, operating STUNner as a public STUN or TURN service still does not go as smoothly as it should.
The upcoming commercial release will allow STUNner to scale to a carrier-grade level. Most importantly, our industry-leading Linux/eBPF based TURN acceleration framework reduces STUNner's CPU footprint more than hundred-fold compared to the default version. This makes it possible to serve beyond 100 Gbps TURN/UDP traffic with a single Linux server. User quota support helps mitigate DDoS attacks, and the STUN server mode, the Daemonset dataplane option, and the relay address discovery features simplify the deployment of STUNner as a public STUN and TURN service. The latter features will be available from the basic member tier, while the full enterprise tier will also include TURN offload on top of all member-tier features. All this with STUNner's unique scalability, security and observability that our users have come to expect throughout the years.
This release contains results of the integration work that will allow us to share the majority of code between the open-source and the commercial branches. Luckily, we could even release some of the features that did not fit into the commercial release in the open source version for free!
Expect the announcement on the first commercial STUNner release in the coming weeks.
Cloud providers' Kubernetes offerings differ widely in terms of WebRTC readiness; especially the UDP LoadBalancer support and the trillions of AWS/EKS subtleties have proved major pain points. Unsurprisingly, lots of users have asked for a reliable way to test and troubleshot a fresh STUNner installation.
The current best practice is to deploy a simple STUNner tutorial and check whether everything works as expected; we recommend the UDP echo service or the "simple-tunnel" tutorials for initial testing. While this is fine for verifying the cases when STUNner actually works, this does not provide useful clues to debug broken installations.
This release contains the first experimental release of the stunnerctl icetest tool. Issuing stunnerctl icetest will deploy a full WHIP server into your Kubernetes cluster, configure STUNner to expose it on a temporary gateway and fire up a WHIP client to connect to the server. If something breaks, the ICE tester tool will print a useful diagnostic message to help fix the problem. Otherwise it will preform a simple load test and report the ICE candidates along with some performance indicators (packet rate, loss, RTT) for both symmetric and asymmetric ICE over UDP as well as TCP.
Note that the ICE test tool is currently experimental: it works most of the time but lacks extensive testing. Please report any issues and help us make this tool a rock-solid helper utility in the upcoming releases.
As usual, this release comes with a number of fixes, additions and improvements all around the place (see the commit logs below). Chiefly among these are new STUNner tutorials (most importantly, a new Elixir example), metric export ported to Open Telemetry, various stunnerctl extensions, and several improvements in the STUNner authentication service.
Enjoy STUNner, join us at Discord, and don't forget to support us!
feature: Add allocation lifecycle event reporting feature: Add license status info to the CDS API feature: Allow stunnerctl to mix JSON queries and strings in the Output Feature: Allow custom user-ids in stunnerctl-auth feature: Extend the CDS client for Node Address Discovery feature: Implement ICE tester in stunnerctl feature: Implement ICE tester server and client feature: Implement license config manager stub feature: Implement per-pod config patcher in the CDS server fix: Allow running with empty auth when setting auth-type=none fix: Fix race condition in logger fix: Fix segfault when NewAuthHandler is called on an empty object fix: Improve CDS server port randomization in tests fix: Make sure listener URI address parses as a valid IP fix: No longer delay config-delete messages in the CDS server refactor: Reimplement ICE tester on top of standard WHIP refactor: Rewrite metric exporting to Open Telemetry chore: Add premium docs to ToC chore: Build and publish icester images chore: Bump Go version to 1.23 chore: Implement quota handler stub chore: Improve error message consistency chore: Make status printer smarter chore: TURN offload integrations chore: Update turncat minimum TLS to 1.2 doc: Add complete public TURN server config doc: Add TURN offload to premium features doc: Document premium features doc: Remove the legacy dataplane mode
feature: Allow offload settings to be configured in the Dataplane feature: Implement Node address patcher in the CDS server feature: Pass the node name to stunnerd in env var feature: Robustify the operator termination codepath feature: Serve the current license status on the CDS server fix: Align Dataplane spec field names with the K8s style guide fix: Clean up lingering dataplane resources (Deployments/DaemonSets) fix: Deprecate the rbac-proxy, fixes l7mp/stunner#185 fix: Generate NoMatchingParent status when route parent is missing Fix: Prevent random restarts by declaring the metrics port in the pod manifest fix: External node IP is preferred over external DNS name refactor: Abstract config renderers behind a generic interface refactor: Add detailed node address discovery logs refactor: Simplify the node controller chore: Add a license manager stub chore: Integrate stub license manager into the rendering pipeline chore: Quota and STUN-mode placeholders chore: Take relay-address placeholder from stunner chore: Update license manager stub from stunner doc: Document that modifying the resource-type in the DP is unsafe test: Robustify tests with CDS server port randomization
feature: Support custom public address in generated configurations fix: Remove the relay-address placeholder from TURN URI addr chore: Add go-generate tags to recreate the OpenAPI bindings chore: Port the auth-service handler to the latest CDS API
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
Nothing published for this version
We are extremely proud to present STUNner v1.0.0, the first stable and generally available release of the STUNner Kubernetes media gateway for WebRTC
We are extremely proud to present STUNner v1.0.0, the first stable and generally available release of the STUNner Kubernetes media gateway for WebRTC brought to you by l7mp.io. This release marks the culmination of a two-year journey to make WebRTC services a first class citizen in the cloud native ecosystem.
There are no major nor minor changes compared to RC5. This release is not without improvements though: we completely reorganized the documentation and introduced two new tutorials, one showing how to deploy STUNner with Elixir and another one for Janus. STUNner now comes with a total of 8 demos, encompassing a wide range of WebRTC applications from cloud-gaming and desktop streaming to video-conferencing. Besides, there are also some minor janitorial fixes throughout the code base.
We started the project a little bit more than two years ago with one simple and ambitious goal: fit WebRTC services natively into the Kubernetes world. Back in those days, the state-of-the-art was to use Kubernetes nodes as mere VMs (the dreaded "host-networking hack"), expose media servers directly to the Internet, and rely on costly 3rd-party STUN/TURN providers for NAT traversal. This model renders essentially all the cloud native selling points ineffective, like elasticity, scalability and manageability, making WebRTC applications an unwelcome guest in the cloud native ecosystem.
Clearly, this had to change. For deploying conventional web apps into Kubernetes it is customary to include a dedicated gateway component with the sole responsibility of exposing the app to the Internet in a secure and scalable way. We wanted to exploit the same pattern for WebRTC, by creating a media gateway for ingesting WebRTC traffic into a Kubernetes cluster over a single IP and port. It just occurred to us that if we implement our gateway on top of the TURN protocol we can also eliminate the need for 3rd party NAT traversal, rendering self-hosting WebRTC services simpler and more economic and cutting down latency significantly. And that is how STUNner was born.
STUNner has come a long way during the two years since its first public release. From a standalone TURN server deployed from a static YAML manifest and customized via environment variables, it has become a fully fledged media gateway service managed by a dedicated Kubernetes operator that implements the official Kubernetes Gateway API. Today STUNner provides many unique features that no other open-source or commercial alternative does, like a fully managed dataplane lifecycle, limitless scale-up/scale-down with thoroughly tested graceful shutdown, and extensive documentation with several detailed tutorials on how to install the most popular WebRTC services into Kubernetes over STUNner.
With this release we consider the project's main goal completed. We have worked closely with our users to help them deploy and troubleshot STUNner with essentially any WebRTC service imaginable, from small-scale to large-scale, from open-source to proprietary, and from proof-of-concept to production stage. No matter which WebRTC application our customers throw at it, STUNner just handles the load reliably, efficiently, and mostly invisibly.
We are new confident that STUNner is ready for the general public and it is time to roll the first GA release.
The v1 release, of course, does not mean an end to our journey.
STUNner has always been an open source software available under a permissive license. The open source development model has proved uniquely fruitful to bring the initial STUNner releases out to early adopters, acquire a solid installed base, and get invaluable bug reports and fixes from our users. This will remain unchanged: STUNner v1 will always be published and maintained as an open-source software.
That being said, the main course of STUNner development will undergo a major change from v1 onward: the future of STUNner development is commercial. We have many exciting new features in preparation, slated to be released soon in the enterprise version of STUNner under an affordable monthly subscription model. Our flagship Linux/eBPF-based UDP/TURN acceleration engine provides millions of packets per second performance per CPU at ultra-low latency. And this is just the start: from Kubernetes operators for deploying and managing full WebRTC applications to Let's Encrypt integration and multi-cloud support, we have many premium features in preparation.
Finally, a round of acknowledgments is in order. First and foremost, we are thankful to our users who supported us from the start, providing us invaluable feedback and early validation of our efforts. Without you, we would have never got to this point. Second, we are grateful to the pion/webrtc project, and especially Sean DuBois: if we didn't have access to a full-featured and an amazingly reusable Golang-based WebRTC toolkit from the outset most probably we would still be struggling with the complexities of libwebrtc desperately trying to release the first usable version of STUNner. Pion is amazing! Last but not least, we thank the work for everyone contributing to our development efforts in any ways: from PRs and bug fixes to documentation updates and bug reports, your help is truly appreciated.
Enjoy STUNner, join us at Discord, and don't forget to support us!
Nothing published for this version
Nothing published for this version
We are proud to present STUNner v0.21.0, the fifth stable release candidate of the STUNner Kubernetes media gateway for WebRTC, brought to you by l7mp
We are proud to present STUNner v0.21.0, the fifth stable release candidate of the STUNner Kubernetes media gateway for WebRTC, brought to you by l7mp.io.
This release candidate marks the next step of the stabilization efforts that will allow us to release the stable v1 version. As such, there have been only minor changes throughout the code base, apart from the usual assortment of documentation fixes, module import updates, and infrastructure upgrades.
In the last release we implemented a minor change to the default load balancer settings, which ended up breaking STUNner for some of our users. This release adds a new annotation that allows to disable this setting for the affected users. Another stability fix (in fact, a one-line commit) makes sure the STUNner gateway operator will safely exit with an error when it fails to start up for some reason (e.g., due to a missing CRD or RBAC policy), instead of just hanging forever. Failing pods early is a safe way to prevent runaway startup crash loops, letting Kubernetes to retry the pod while enforcing a secure backoff algorithm.
By default STUNner applies the sessionAffinity: ClientIP setting on the LB services it creates to expose Gateways. Normally this setting improves stability by ensuring that each TURN session is safely pinned to the same dataplane pod for its entire lifetime. Certain hosted Kubernetes platforms, however, seem to reject UDP LB services that have this setting on, breaking STUNner deployments on these systems.
In order to prevent STUNner from enforcing session affinity on the LB Service corresponding to a Gateway, you can now set the stunner.l7mp.io/disable-session-affinity: true annotation on the Gateway. Otherwise, session affinity is turned on.
fix: Add annotation to unset LB Service session-affinity, fix l7mp/stunner#155
fix: Exit when the operator fails to start up
doc: Document the disable-session-affinity:true annotation
chore: Add buildinfo to the stunnerctl binary
chore: Bump pion/turn to v4 and update some deps
chore(rtd): Update rtd dependencies
Enjoy STUNner and don't forget to support us!
Nothing published for this version
Nothing published for this version
Nothing published for this version
One column per quarter.