Akka insights

Skill vezril/claude-toolkit/skills/akka-insights

Akka Insights (Cinnamon, formerly Lightbend Telemetry) — intelligent monitoring and observability purpose-built for Akka. A JVM agent collects metrics, events, and traces from Akka internals (actors, mailboxes, dispatchers/thread pools, remoting, cluster, sharding, persistence, streams, projections, HTTP, gRPC, Futures) and exports them to backends (Prometheus, OpenTelemetry, Datadog, StatsD, Elasticsearch) with Grafana/Datadog dashboards and Jaeger/Zipkin tracing. Covers how it's added (sbt/Maven plugin + javaagent + commercial credentials), what it instruments, the metrics/events/traces model, backends and visualizations, and that it is a commercial/licensed product. Use when setting up production monitoring/observability for an Akka application, diagnosing actor/cluster/sharding/persistence/stream performance, or choosing what Akka metrics to watch. Commercial; complements akka-diagnostics.From its SKILL.md

Install

npx -y skills add vezril/claude-toolkit --skill akka-insights

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.3 KB, as published. Nobody here has run it

Akka Insights (Cinnamon)

Intelligent monitoring/observability purpose-built for Akka (a.k.a. Cinnamon, formerly Lightbend Telemetry). A JVM agent in your application process collects metrics, events, and traces from Akka internals and exports them to monitoring backends with ready-made dashboards. It is a commercial / licensed product (requires an Akka subscription). Complements the free [[akka-diagnostics]] (which is startup-time checks, not continuous telemetry).

Cross-links: [[akka]] (meta), [[akka-diagnostics]], [[akka-cluster]], [[akka-persistence]], [[akka-streams]].

How it's added

A Cinnamon Agent (-javaagent) attaches at JVM startup and instruments based on your config. In a cluster it runs on each node, each reporting independently. Setup (sbt example; Maven/Gradle equivalent):

// project/plugins.sbt — needs Akka commercial resolvers/credentials
addSbtPlugin("com.lightbend.cinnamon" % "sbt-cinnamon" % "2.22.x")
// build.sbt
lazy val app = (project in file(".")).enablePlugins(Cinnamon)
run / cinnamon := true
test / cinnamon := true
libraryDependencies ++= Seq(
  Cinnamon.library.cinnamonAkka,          // instrument Akka (required)
  Cinnamon.library.cinnamonAkkaTyped,
  Cinnamon.library.cinnamonAkkaPersistence,
  Cinnamon.library.cinnamonAkkaCluster,
  Cinnamon.library.cinnamonAkkaStream,
  Cinnamon.library.cinnamonPrometheus)    // a backend (required to export)

Both cinnamonAkka and a backend (cinnamonPrometheus/cinnamonCHMetrics/…) are required to instrument Akka. Actor telemetry is off by default (apps may have tens of thousands of actors) — opt in per actor path in application.conf:

cinnamon.akka.actors { "/user/*" { report-by = class } }

What it instruments & the model

Instrumentations: Akka Actors, Remoting, Cluster, Persistence, Cluster Sharding, Streams, Projections, HTTP (server/endpoint/client), gRPC, and Scala/Java Futures.
Three instrument kinds: Metrics (counters/gauges/rates — actors, mailbox sizes, dispatcher/thread-pool utilization, cluster membership, shard counts, persistence latencies, stream throughput, HTTP latencies), Events (errors, unhandled messages, dead letters), Traces (follow async/distributed message flows). Extensions add OpenTracing/SLF4J-MDC context propagation, custom metrics/events, and JMX/JVM importers.
Backends (exporters, multiple can run at once): OpenTelemetry, Prometheus, Datadog, Coda Hale Metrics, StatsD, Elasticsearch, SLF4J events, Telegraf; tracing reporters Jaeger, Zipkin, Datadog. Visualizations: Grafana and Datadog dashboards, Vizceral.
A developer sandbox (Elasticsearch+Kibana+Grafana, or OpenTelemetry/Prometheus variants) lets you try it without existing monitoring infra.

Always-apply defaults

It's commercial — needs an Akka subscription and the Akka commercial resolvers/credentials (don't commit credential URLs). For free checks (config/starvation) use [[akka-diagnostics]] instead; for raw metrics you can also use the Core libraries' own metric hooks.
Enable cinnamonAkka + one backend; add per-toolkit modules (persistence/cluster/sharding/streams/http) matching what you run.
Don't instrument every actor — actor telemetry is off by default; select meaningful paths (report-by = class to group) to avoid metric explosion.
Watch the Akka-specific signals that matter operationally: dispatcher/thread-pool saturation and mailbox growth (ties to [[akka-diagnostics]] starvation), cluster membership/unreachable, shard distribution, persistence/recovery latency, stream backpressure, and HTTP latency.
Export to your existing stack (Prometheus/Grafana or OpenTelemetry) and add tracing (Jaeger/Zipkin) for cross-node message flows.

Anti-patterns (flag in review)

Instrumenting all actors (metric/cardinality explosion); shipping without a backend module (nothing exported).
Treating Insights as a substitute for [[akka-diagnostics]]' config/starvation checks (they're complementary); ignoring mailbox/dispatcher metrics that signal blocking.

[[akka-diagnostics]] (free startup checks) · [[akka-cluster]] / [[akka-persistence]] / [[akka-streams]] (the subsystems it instruments) · [[akka]] (meta).
Source: https://doc.akka.io/libraries/akka-insights/current/