netdata/netdata
The fastest path to AI-powered full stack observability, even for lean teams.
How netdata is put together
Netdata is a real-time infrastructure monitoring agent that collects per-second metrics on Linux, macOS, FreeBSD, and Windows. The core daemon is written in C (built with CMake), with data-collection plugins implemented as separate processes in Go (src/go) and Rust (src/crates), communicating with the agent over plugin IPC protocols. Data flows from collectors through the agent's metric engine into its embedded time-series database with tiered storage, plus ML-based anomaly detection, alerting, streaming to parent nodes, and the ACLK cloud connection (src/aclk). Packaging spans static makeself bundles, Docker images, native RPM/DEB repos, and a Windows MSI installer.
Languages
Frameworks
Datastores
Infrastructure
Major components
Core agent daemon (C)
Main netdata process built by the root CMakeLists.txt: metric collection orchestration, storage engine, health/alerting, web API, and streaming.
ACLK (Agent-Cloud Link)
src/aclk implements the MQTT/WebSocket connection to Netdata Cloud, including alarm and contexts APIs.
Go collectors/plugins
src/go contains the go.d.plugin-style collectors (SNMP, Kubernetes, Docker, databases, Prometheus scrape, etc.) plus cgroups.plugin helpers like cgroup-name.
Rust crates workspace
src/crates hosts journal/log indexing engines, OTLP ingest/indexing, WAL, chunk-file storage format, NetFlow/sFlow parsing, and the netdata-plugin IPC bridge crates.
Packaging & distribution
packaging/ provides kickstart/static installers, makeself bundles, Docker images, DEB/RPM repo config, Windows MSI build scripts, and Dagger-based CI pipelines (packaging/dag).
Integrations catalog generator
integrations/ holds Python generators and JSON schemas that produce documentation and metadata for collectors, exporters, notifications, and secret stores.
agent-events ingestion tool
packaging/tools/agent-events/server.go is a standalone HTTP deduplication/event-ingestion service exposing metrics, health, and OTLP log export.
Installer & system integration
netdata-installer.sh, system/, and packaging/installer handle installation, updater, uninstaller, and service configuration across platforms.
Query engine correctness fixes dominated, along with retiring the legacy HAProxy collector and a batch of dependency updates.
5 weeksWeek of 2026-08-24
Week of 2026-08-24latest
Query engine correctness fixes dominated, along with retiring the legacy HAProxy collector and a batch of dependency updates.
Week of 2026-08-17
Ceph monitoring got major attention with new dashboards and alerts, plus eBPF code was converted to Go and SNMP discovery improved.
Week of 2026-08-10
Version 2.11.0 shipped, with new Prometheus profiles for AI inference and storage systems, plus stronger TLS verification for cloud connections.
Week of 2026-08-03
Prometheus profile capabilities expanded with relabeling and fallback types, while security hardening touched the installer and logging paths.
Week of 2026-07-27
Operational testing infrastructure grew substantially across collectors, and the SNMP traps collector was refactored for cleaner ownership.
Over the past five weeks, Netdata's development focused heavily on expanding and refining its Prometheus-based monitoring profiles (Ceph, vLLM, LiteLLM), hardening the query engine with a suite of correctness fixes, and improving SNMP device discovery and topology reporting. A minor release (v2.11.0) shipped mid-August, alongside security improvements like TLS hostname verification for cloud connections and safer string handling. The team also retired the old HAProxy collector in favor of the Go-based one, converted eBPF algorithms to Go, and invested in operational testing infrastructure for collectors.
Week by week
2026-08-24Query engine correctness fixes dominated, along with retiring the legacy HAProxy collector and a batch of dependency updates.latest6 changes
Fix
HAProxy collectors retired
Both the Python-based and dedicated Go HAProxy collectors were removed, consolidating on the standard Go collector.
Fix
Query engine accuracy fixes
A series of fixes ensured sums, counts, timestamps, and baselines are preserved correctly across data boundaries when querying metrics.
Fix
Safer WebSocket handling
Fixed pooled response buffers and fragmented WebSocket payload masking to prevent data corruption.
Feature
Palo Alto BGP support
The PAN-OS collector can now gather BGP peer information from devices using the Advanced Routing Engine.
Fix
SNMP UniFi model detection
UniFi network devices are now correctly identified from their system descriptions.
Refactor
Faster chart metadata handling
Removed redundant re-processing of chart metadata to speed up metric storage.
2026-08-17Ceph monitoring got major attention with new dashboards and alerts, plus eBPF code was converted to Go and SNMP discovery improved.6 changes
Feature
Ceph monitoring coverage
Added comprehensive Ceph storage cluster monitoring charts and alerting rules as part of a phased rollout.
Refactor
eBPF algorithms moved to Go
DC (dirty page tracking) algorithms were rewritten in Go for the eBPF plugin, replacing older implementations.
Chore
Query engine test corpus
Introduced layered black-box tests that verify the query engine produces correct results across many scenarios.
Fix
SNMP topology accuracy
Fixed how discovered devices are counted and reported so network topology views show correct metadata.
Feature
Windows network viewer support
The network-connections view now works on Windows machines.
Fix
Memory overhead reduction
Reduced the performance cost of memory usage tracking on Linux systems.
2026-08-10Version 2.11.0 shipped, with new Prometheus profiles for AI inference and storage systems, plus stronger TLS verification for cloud connections.6 changes
Chore
Minor release v2.11.0
The team cut the v2.11.0 release after weeks of accumulated fixes and features.
Feature
New monitoring profiles
Added holistic monitoring profiles for vLLM and LiteLLM (AI inference servers) and Ceph (storage).
Fix
Stronger cloud connection security
HTTPS requests to the cloud now verify both the server's hostname and IP address over TLS.
Refactor
Ceph collector reworked
The Ceph collector was rebuilt to complement Prometheus metrics rather than duplicate them, and migrated to the newer framework.
Fix
macOS disk health fixes
Fixed SMART disk health checks on macOS by correcting IOKit paths and memory lifecycle issues.
Docs
OpenTelemetry documentation
Published new documentation explaining Netdata's OpenTelemetry integration.
2026-08-03Prometheus profile capabilities expanded with relabeling and fallback types, while security hardening touched the installer and logging paths.6 changes
Feature
Prometheus metric relabeling
Profiles can now rename and transform incoming Prometheus metrics before they're stored.
Fix
Installer security fix
Sanitized arguments to the kickstart installation script so they can't be misused in privileged contexts.
Feature
eBPF and network viewer integration
Connected the eBPF plugin with the network viewer plugin so low-level network events feed into network views.
Fix
Logging format safety
Validated host name prefixes and replaced risky format strings to prevent crashes or injection via printf-style functions.
Feature
Chart instance labels
The charting engine now supports optional per-instance labels for more flexible dashboard grouping.
Fix
Spawn server reliability
Fixed the listen backlog for the process-spawning server and added a test to keep it working.
2026-07-27Operational testing infrastructure grew substantially across collectors, and the SNMP traps collector was refactored for cleaner ownership.6 changes
Chore
Operational tests for secrets and configs
Added real-world tests verifying Vault secret storage, HTTP configuration discovery, AWS secrets, and dynamic configuration all work end-to-end.
Refactor
SNMP traps cleanup
Refactored the SNMP traps collector across several PRs to internalize output backends and reduce its exposed surface.
Feature
Cloudwatch configurability
The AWS CloudWatch collector can now be configured directly from the dashboard UI, with per-selection query policies.
Refactor
NetFlow performance
Added capacity benchmarks and optimized shared ingestion paths for processing NetFlow network traffic data.
Fix
NetFlow filtering fixes
Corrected how NetFlow filters work and how search facets are retained.
Fix
Security fix in URL decoding
Prevented an out-of-bounds read vulnerability in URL percent-decoding.
Dependencies and code review
Dependency advisories
Security Watch
No known advisories across 0 scanned dependencies.
No known advisories in the scanned dependencies.
Code review
No concrete code-level issues confirmed.
Get this report every week for your repos.
GitZoid learns each repo, reports what changed, and flags what needs attention. One flat price for the whole team.
$19 a month, flat · First 10 outputs free · No card required