Your observability stack shouldn't have more dependencies than your app.

Every external service you add is another thing that can break. aacyn is one binary — deploy it, and you're done.

Dependency chains: aacyn vs. the competition

Count the boxes. Fewer boxes = fewer failure modes, fewer alerts, less YAML to maintain.

aacyn

Self-contained. Deploy the binary, get everything.

Helm installDone

Single binary — no external database, no time-series store, no query engine. The C columnar store handles ingest, scans, and serving in one process.

Coroot

Open-source, but requires 3 external services.

Helm installClickHouseGrafanaPrometheusDone

ClickHouse for storage, Prometheus for metrics, Grafana for dashboards. Each dependency adds operational overhead and a failure domain.

Pixie

Great eBPF, but ties you to New Relic.

Helm installIn-cluster storeNew RelicDone

In-cluster storage handles short-term data. Long-term retention and dashboards require New Relic Cloud or a self-hosted alternative.

Grafana Beyla

eBPF with Grafana's ecosystem requirements.

Helm installGrafanaPrometheus / MimirDone

Beyla emits metrics and traces; you still need Grafana for dashboards and Prometheus or Mimir for storage. Each layer must be deployed and maintained separately.

Cilium Hubble

Powerful, but requires Cilium CNI migration.

Migrate CNI to CiliumHubble UIDone

Hubble only works on Cilium. Migrating a production CNI is a multi-hour, high-risk operation. If you're not already on Cilium, Hubble is not an option.

Inspektor Gadget

Flexible toolkit, but not a unified platform.

Install gadgetPer-gadget configExternal dashboardDone

Each gadget is a standalone tool. No unified topology view, no built-in golden signals, no pre-configured alerts. Assembly required.

1binary to deploy
0external databases
30sto first golden signal