Skip to content

SigNoz

Summary

SigNoz is an open-source, OpenTelemetry-native observability platform. It puts logs, metrics, traces, APM, dashboards, alerts, exceptions and (since v0.143.0) AI/LLM observability in one UI, backed by ClickHouse. It runs as three pieces: a single signoz binary (UI, API, ruler, Alertmanager, OpAMP), an OpenTelemetry Collector distribution (the ingester), and ClickHouse plus a small SQLite or PostgreSQL metastore. The core is MIT-licensed, enterprise features live in ee/ under a commercial license, and the collector is AGPL-3.0. Self-hosted installs now go through Foundry (foundryctl) or the Helm chart. SigNoz Cloud starts at $49/month with usage-based pricing.

Key Facts

Attribute Detail
Latest Version v0.143.0 (2026-09-23). Needs signoz-otel-collector v0.144.11 (2026-09-22).
Release cadence Weekly minor releases (Wednesdays) plus patch releases as needed. Still pre-1.0.
Repository github.com/SigNoz/signoz
License MIT (Expat) outside ee/ and cmd/enterprise/. SigNoz Enterprise License for ee/. Collector: AGPL-3.0.
Company SigNoz Inc. (Y Combinator W21). About $6.5M seed funding led by SignalFire (announced September 2023).
Language Go (backend, collector), TypeScript/React (UI)
Storage ClickHouse for telemetry. SQLite (binary default) or PostgreSQL (Foundry default) for metadata.
Install paths Foundry (Docker Compose, Swarm, systemd, Kubernetes Helm/Kustomize, ECS, Coolify, Railway, Render), Helm chart signoz/signoz 0.143.0, SigNoz Cloud
Cloud pricing $49/month base (includes $49 of usage). $0.30/GB logs and traces. $0.10 per million metric samples. No seat or host fees.
Compliance (Cloud) SOC 2 Type II and HIPAA (per README)

Full version, port, schema and pricing tables are in Reference.

Architecture at a Glance

Data flows from OpenTelemetry sources through the SigNoz collector into ClickHouse. The single SigNoz binary queries it and manages the collector over OpAMP:

flowchart LR
    APPS["Apps + OTel SDKs<br/>k8s-infra agents"] -->|"OTLP :4317/:4318"| ING["ingester<br/>signoz-otel-collector"]
    ING -->|"clickhousetraces, clickhouselogsexporter,<br/>signozclickhousemetrics"| CH[("ClickHouse<br/>+ Keeper/ZooKeeper")]
    SZ["signoz :8080<br/>UI, API, ruler,<br/>Alertmanager, OpAMP"] -->|SQL| CH
    SZ --> META[("Metastore<br/>SQLite / PostgreSQL")]
    SZ -.->|"OpAMP config"| ING
    USERS["Users, API keys,<br/>MCP clients"] --> SZ

The full internals (ingestion pipeline, storage layout, migrator, query path, security model) are in Explanation.

Evaluation

Why it stands out

  • OpenTelemetry-native. There is no proprietary agent. Any OTLP source works, and the collector bundles over a hundred upstream receivers.
  • One store, correlated signals. Traces, logs and metrics live in ClickHouse, so you can move between them by trace_id and service without leaving the UI.
  • Predictable pricing. No per-seat, per-host or custom-metric charges on Cloud. Self-hosted Community is free.
  • Fast-moving AI features. v0.143.0 added AI Observability (token, cost and latency views, gen_ai.* normalization, per-model pricing). There is also an MCP server, and Noz, an AI assistant that is Cloud only.

When it fits

  • Teams standardizing on OpenTelemetry who want one UI instead of Prometheus + Loki + Tempo/Jaeger.
  • Teams leaving Datadog or New Relic for cost reasons that can run ClickHouse, or that choose SigNoz Cloud.
  • High-cardinality log and trace workloads, where ClickHouse's columnar storage beats inverted-index log stores.
  • LLM and agent applications that already emit gen_ai.* spans.

Pros and cons

Pros Cons
OTel-native, no proprietary agents Self-hosting means running ClickHouse, Keeper/ZooKeeper and schema migrations
One UI for traces, logs, metrics, alerts, exceptions Pre-1.0 weekly releases with breaking changes and required upgrade stops
ClickHouse: fast aggregations, good compression SAML/OIDC SSO, custom roles and anomaly alerts need Enterprise or Cloud
Transparent usage pricing ($0.30/GB logs and traces) OTLP ingestion is unauthenticated on self-hosted. You must secure it yourself.
AI/LLM observability, MCP server No continuous-profiling signal in the product (as of 2026-09)
Foundry: one casting file for many platforms Deploy tooling changed in 2026 (Compose files deprecated in v0.130.0)
Frontend/browser telemetry and Web Vitals via OTel Smaller dashboard/plugin ecosystem than Grafana

Key Features

Feature Detail
APM Service latency, error rate, throughput, Apdex, top endpoints, DB and external calls (from signozspanmetrics)
Distributed tracing Flamegraph and waterfall, span events, trace analytics, trace funnels
Logs OTel logs, UI-managed log pipelines pushed over OpAMP, query builder, trace correlation
Metrics and dashboards Query builder, PromQL, ClickHouse SQL. Text, heatmap and area panels.
Infrastructure Kubernetes clusters, nodes, pods, containers (v0.140.0), hosts
Alerts Metric, log, trace and exception rules. Anomaly alerts (Cloud/Enterprise). Slack, PagerDuty, Opsgenie, MS Teams, Jira, incident.io, webhook, email.
AI/LLM observability Overview, AI explorer, attribute mapping, per-model pricing (v0.143.0). 55 documented provider and framework integrations.
Agent-native SigNoz MCP server (Cloud-hosted or self-deployed on port 8000). Noz assistant (Cloud only).
Frontend monitoring Web Vitals and browser telemetry through OpenTelemetry JS
Access control Service accounts and API keys, SSO, managed roles. Custom roles are in beta.

Licensing and Pricing

  • Community (self-hosted): free, MIT core, no feature gating on core signals.
  • Enterprise Self-Hosted: code in ee/ needs a paid license in production. Development and testing are allowed without one.
  • SigNoz Cloud Teams: $49/month base fee that covers $49 of usage (it used to be $199). Then $0.30/GB for logs and traces and $0.10 per million metric samples. 30-day free trial. Default retention is 15 days for logs and traces and 30 days for metrics.
  • Enterprise Cloud / BYOC: from $4,000/month, which includes $4,000 of monthly usage; volume discounts and custom retention by contract (SigNoz pricing, checked 2026-09-28).

See the pricing table and edition matrix.

Compatibility

Dimension Support
Ingestion OTLP gRPC/HTTP by default. Collector bundles Jaeger, Zipkin, Prometheus scrape, Prometheus remote-write, FluentForward, syslog, Kafka, Loki, Datadog, Splunk HEC and more. You must enable them in the config.
Query languages Query builder, PromQL, ClickHouse SQL
Storage ClickHouse only (Foundry and the chart ship 25.12.5)
Deployment Foundry (Docker Compose/Swarm, systemd, Kubernetes, ECS, PaaS), Helm, SigNoz Cloud
Architectures amd64, arm64 (collector and SigNoz release tarballs for both)
Hosts Linux, macOS. Windows only through WSL 2.
Kubernetes Chart README lists Kubernetes 1.16+ and Helm 3.0+

Alternatives and Migration

Alternative Choose it instead when
LGTM Stack You want object-storage-backed, per-signal scaling and the Grafana ecosystem
Victoria Stack Metrics and logs efficiency on local disks matter most
OpenObserve You want a single binary on object storage
Coroot You want eBPF auto-instrumentation and profiling
Apache SkyWalking You need an Apache-governed APM with its own agents
ClickStack (ClickHouse + HyperDX) You want ClickHouse Inc.'s own OTel stack. See Observability 2.0.

Lock-in: low at the instrumentation layer, because everything is OTLP. Dashboards, alerts and saved views are SigNoz-specific, though a Terraform provider (signoz/signoz) lets you keep them as code. Prometheus recording rules have no direct equivalent. Re-create them as alerts or ClickHouse materialized views.

Topic Map

  • How-to Guides: install with Foundry or Helm, migrate off Compose, upgrade, retention, scaling, collector customization, security, day-2 commands (v0.143.x).
  • Reference: versions, editions, pricing, components and ports, ClickHouse tables, retention defaults, config keys, Helm values, capacity, benchmarks, hardening checklist.
  • Explanation: the SigNoz binary, OpenTelemetry ingestion, ClickHouse storage design, schema migrator, query execution, alerting, security model, trade-offs.

Sources

Source URL
Website and docs https://signoz.io/docs/
Architecture https://signoz.io/docs/architecture/
Repository and README https://github.com/SigNoz/signoz
Release v0.143.0 https://github.com/SigNoz/signoz/releases/tag/v0.143.0
Upgrade guide v0.143.0 https://signoz.io/docs/operate/migration/upgrade-0-143/
Upgrade path and release cycle https://signoz.io/docs/operate/upgrade/
License / Enterprise license https://github.com/SigNoz/signoz/blob/main/LICENSE · https://github.com/SigNoz/signoz/blob/main/ee/LICENSE
SigNoz OTel Collector https://github.com/SigNoz/signoz-otel-collector
Foundry https://github.com/SigNoz/foundry
Helm charts https://github.com/SigNoz/charts
Docker install (Foundry) https://signoz.io/docs/install/docker/
Compose-to-Foundry migration https://github.com/SigNoz/signoz/blob/main/deploy/MIGRATION.md
Resources planning https://signoz.io/docs/setup/capacity-planning/community/resources-planning/
Pricing https://signoz.io/pricing/
$49 Teams plan announcement https://signoz.io/blog/cloud-teams-plan-now-at-49usd/
LLM observability https://signoz.io/docs/llm-observability/
Funding announcement https://signoz.io/blog/signoz-funding/
Logs benchmark https://signoz.io/blog/logs-performance-benchmark/
Changelog https://signoz.io/changelog/

Questions

Open Questions

  • Which exact release merged query-service and frontend into the single signoz binary? The architecture doc dated 2025-04-03 already describes the bundled binary.
  • Is running several signoz replicas supported for HA in Community, given that the sharder defaults to noop?
  • Is there an up-to-date, independently reproduced benchmark for the current logs_v2 / signoz_index_v3 schemas?
  • How does SigNoz roll back a bad OpAMP-pushed collector config? The protocol allows it, but SigNoz's own behaviour is undocumented.

Answered Questions

  • What does Enterprise Cloud cost? From $4,000/month, including $4,000 of monthly usage (SigNoz pricing, checked 2026-09-28).
  • Is the core open source or source-available? Open core. Everything outside ee/ and cmd/enterprise/ is MIT. ee/ uses the SigNoz Enterprise License, which requires a paid subscription for production use and allows free dev/test use. The collector is AGPL-3.0. Sources: LICENSE, ee/LICENSE, collector README.
  • How does SigNoz handle ClickHouse schema migrations? The signoz-otel-collector migrate subcommands ready, bootstrap, sync up and async up run as a one-shot migrator job. Collectors run migrate sync check before they start. Replication-aware DDL is on by default in Helm (telemetryStoreMigrator.enableReplication: true). See Explanation.
  • Maximum tested ClickHouse cluster size? Not published. The docs show 2 shards x 2 replicas with 3-node ZooKeeper as the reference, and the resources guide sizes 2 shards at 16 CPU / 32 GiB each.
  • Can SigNoz ingest Prometheus recording rules directly? No. Queries and alerts support PromQL, and metrics arrive through the prometheus or prometheusremotewrite receivers, but you must re-create recording rules as alerts or ClickHouse materialized views.
  • What storage backend does SigNoz use? ClickHouse only for telemetry, plus SQLite or PostgreSQL for metadata. See Architecture at a Glance.
  • Does SigNoz support PromQL? Yes, for metrics dashboards and alerts, alongside the query builder and ClickHouse SQL.
  • How does it handle high cardinality? Attributes are stored in typed Map columns and can be promoted to materialized columns, with no inverted index. Metric label cardinality still creates series. Optional reduction rules can aggregate labels away at write time. See Explanation.
  • Which cloud instance types are recommended? AWS T3/T4g for general components and C5/C6g/C7g for ClickHouse. GCP E2 and C3/C3D. Azure Dsv5 and Fsv2/Fasv6. See Reference.
  • Does SigNoz support RUM / frontend monitoring? Yes. Web Vitals and browser telemetry go through OpenTelemetry JS.
  • Is Docker Compose still supported? The bundled Compose files and install.sh were deprecated in v0.130.0. Foundry generates Compose files instead. See How-to Guides.
  • Do self-hosted installs need ingestion keys? No. Ingestion keys are Cloud-only, and self-hosted OTLP is unauthenticated by default.