Your app was slow for six hours. Uptime Kuma said everything was green — because the homepage returned 200. The checkout API was timing out, logs were scattered across three containers, and nobody could trace a single user request from nginx to Postgres to Redis. Datadog would fix it. Datadog would also invoice you like a second mortgage.
SigNoz is the self-hosted middle ground. ~27k GitHub stars, open source, OpenTelemetry-native — logs, metrics, and distributed traces in one UI. ClickHouse backend. Flamegraphs, service maps, p99 latency, log search, dashboards, alerts. The Datadog-shaped toolbox without per-host SaaS math.
What it actually does
Instrument your apps with OpenTelemetry SDKs (Python, Go, Node, Java, .NET, Ruby, and more). Telemetry flows to SigNoz's bundled OpenTelemetry Collector on ports 4317 (gRPC) and 4318 (HTTP). SigNoz stores and queries everything in ClickHouse.
APM. p99 latency, error rates, Apdex, ops/sec out of the box. See database calls and external HTTP requests per service — which dependency is actually slow.
Distributed tracing. Follow a user request across microservices with flamegraphs and Gantt charts. Click a slow span, jump to correlated logs. This is what Uptime Kuma can't do — Kuma tells you the site is up; SigNoz tells you why checkout took 8 seconds.
Logs. Centralized log management with fast search and query builder. Create charts from log patterns, pin them to dashboards.
Metrics. Custom dashboards — time series, pie charts, bar charts. PromQL support plus a visual query builder.
Alerts. Threshold alerts on logs, metrics, or traces. Slack, PagerDuty, webhooks — catch the spike before customers tweet about it.
LLM observability. Token usage, latency, prompt/response debugging for AI apps. Overlaps somewhat with Langfuse, but SigNoz is the full-stack ops platform; Langfuse is laser-focused on LLM engineering workflows.
SigNoz vs what we've already covered
- Uptime Kuma — synthetic uptime checks and status pages
- Langfuse — LLM traces, prompt versions, eval datasets
- ClickHouse — the database; SigNoz runs on top of it for observability specifically
SigNoz is infrastructure and application observability for teams running real services — APIs, workers, queues, databases. You might run Uptime Kuma and SigNoz: external synthetic checks plus internal request tracing.
Why self-host?
Telemetry is sensitive. Traces contain user IDs, SQL queries, request paths, error stack traces. Logs might include PII. Shipping that to a US observability vendor is a compliance conversation you don't want at 4pm Friday.
Predictable costs. Datadog and New Relic bill per host, per GB ingested, per feature unlocked. Self-hosted SigNoz costs whatever your VPS runs — flat monthly, no surprise invoice when log volume doubles.
OpenTelemetry, no lock-in. OTel is the instrumentation standard. Switch collectors or backends later without rewriting every service. SigNoz embraces that; proprietary agents don't.
Canadian data residency. Run the whole stack — collector, ClickHouse, UI — on Montreal or Toronto hardware. Your traces never leave the country.
What running it takes
Docker Compose is the standard install:
git clone -b main https://github.com/SigNoz/signoz.git
cd signoz/deploy/docker
docker compose up -dUI on port 8080. Stack includes ClickHouse, ZooKeeper, OTel Collector, and the SigNoz query service. Plan 2–4 GB RAM at idle; production ingest with high trace volume wants more CPU, RAM, and fast SSD for ClickHouse.
Point apps at your collector:
OTEL_EXPORTER_OTLP_ENDPOINT=http://your-server:4318
OTEL_SERVICE_NAME=my-api
OTEL_TRACES_EXPORTER=otlp
OTEL_METRICS_EXPORTER=otlp
OTEL_LOGS_EXPORTER=otlpKubernetes deployments use their Helm chart. Set retention TTLs on ClickHouse or disk grows forever. Put auth and TLS in front of the UI if it's not VPN-only.
Who it's for (and who should skip it)
Good fit: dev teams with microservices or multi-container apps, agencies hosting client APIs in Canada, anyone who outgrew grep-and-pray log diving, shops allergic to Datadog bills.
Maybe skip it: one static WordPress site with no custom backend — Uptime Kuma is enough. Solo dev with one process and twelve users — structured logging to a file might suffice until pain arrives. SigNoz pays off when you have services talking to services and need to correlate signals.
Hosting it in Canada
SigNoz is a RAM-and-disk hunger game. We deploy it on Canadian Docker hosting — sized for ClickHouse growth, OTLP ingestion, backups on the telemetry store, and network placement with your apps.
Tell us your service count and log volume — we'll size the box before ClickHouse eats the partition during your first traffic spike.
