What netdata/netdata shipped

This week: Cloud queries and parallel weights moved off the libuv pool; out-of-order health saves no longer drop a parent's alerts; SNMP interface charts now carry ifAlias as a label; FreeBSD gained missing memory and battery metrics; ppc32, mips32 and armv5 builds fixed.

Changes for Netdata users this week: Cloud data queries and parallel weights chunks no longer run as libuv pool jobs, so a weights query over many nodes or a burst of Cloud queries can no longer occupy every pool thread. Health keeps an alert's latest transition when saves arrive out of order, fixing parents that showed no alerts in Netdata Cloud while still having them raised. SNMP collects ifAlias as the if_alias label on interface charts, so ports can be filtered and grouped by the description an administrator assigns instead of the raw interface name. Stock profile OIDs were also corrected against MIBs and device captures. FreeBSD gains missing ACPI battery capacity and committed memory metrics, diskspace now charts a filesystem whose kept mount is excluded, and ppc32, mips32 and armv5

This week: fixed the dbengine datafile use after free behind the DEADLOCK DETECTED fatal, keyed the page cache on metric ids, held chart references in context walks, capped chunked ACLK responses at 1 MiB, and put HTTP response bodies under a 64 MiB ceiling.

This week's work on Netdata: a use after free in dbengine that produced a long-standing fatal error at startup is fixed, and the page cache now keys on the metric id rather than a memory address. Context readers now hold a chart reference while reading instance labels and walking instance charts. On the network side, ACLK chunk sizes are strictly parsed and chunked responses capped at 1 MiB, and shared HTTP response bodies are bounded at 64 MiB. Network Flows and OpenTelemetry traces now require a paid Cloud space.

Netdata v2.12.2 is out. This week also fixes a streaming crash on connect, host removal races, Kubernetes charts for similarly named objects, and SQLite corruption detection. Plus more Exchange metrics on Windows and Rust tests in CI.

Netdata v2.12.2 is a patch release that tightens SNMPv3 authentication in the SNMP traps receiver, stops alert evaluation from keeping obsolete charts alive, and makes Cloud see disconnected children as offline. Alongside it, this week brings a streaming fix that ends a crash on every connect, host removal that no longer frees a host still in use, Kubernetes chart cleanup that only removes the deleted object's charts, earlier detection of SQLite metadata corruption, payload framing checks for function calls, and eBPF plugins that use compiled-in directories when privileged. Windows gains more Exchange metrics, and Rust code changes now get cargo test in CI.

This week: go.d alternative configs for same-name jobs, otel-streams websocket reconnect, the 125% child offline wait now actually applied, offline reported only after the online flag clears, reconnect backoff clamps fixed, and static aarch64 agents on Cortex-A53 no longer crash-

This week's work on Netdata covered several reliability fixes across the agent and parent. go.d configurations sharing a collector and job name can now serve as alternative connection attempts, and otel-streams load generators recover from websockets that go silent. The parent again waits 125% of a child's reboot time before reporting it offline, and only reports after the online flag is cleared. Reconnect backoff clamps return milliseconds, and MQTT packet id 1 is no longer handed out twice at counter wrap. Re-created charts keep the instance they took over, and alert evaluation no longer keeps obsolete charts alive. Static aarch64 agents on Cortex-A53 boards work around two errata that caused SIGILL crash loops, and a new host label exposes the Jetson Linux release.

Two releases this week: v2.12.1 fixes the Windows Agent startup crash and timezone bug, stabilizes eBPF across more kernels, and cuts Docker collector load. Also: Go IPMI now covers all Linux architectures and remote LAN BMCs, and SNMP no longer logs secrets.

This week Netdata shipped v2.12.1, a patch release that fixes a Windows Agent startup crash and a timezone detection bug, stabilizes the eBPF collector across a wider range of kernels and libbpf versions, and reduces the load the Docker collector places on the system. v2.12.0 was also published, with its Windows installer removed from the release page because of startup crashes traced to an upstream bug. On the collectors side, the experimental Go IPMI plugin now supports all Go Linux architectures and can collect from remote BMCs over LAN and LAN+, the Jetson collector added per-rail power, and SNMP stops putting communities and v3 passphrases in debug logs. Host labels now reload from the config file the Agent loaded, and DEM gained Agent-aware GeoIP refresh and explicit RUM installation

Netdata v2.12.0: OpenTelemetry Traces, Playwright journeys and Lighthouse audits in dem.plugin, Redfish hardware monitoring, Redis-compatible servers, and Windows Cluster Shared Volumes support.

Netdata v2.12.0 ships with OpenTelemetry Traces in Early Access, enabling observability across tracing, metrics, and logs. dem.plugin now runs Playwright journeys and Lighthouse audits natively. Redfish brings out-of-band hardware monitoring. New support for Redis-compatible servers (Valkey, Dragonfly, KeyDB, Kvrocks, Garnet). Windows gains Hyper-V, RDP, and Cluster Shared Volumes monitoring. The native scripting system now eliminates separate metadata files.

New: StatsD runs as statsd.d.plugin. Fixes: function response serialization, database flush deadlock, idle CPU on 32-bit time64, SMBIOS slot identification, host label streaming. zstd archives now available.

This week's updates strengthen Netdata's reliability across multiple fronts. StatsD is now a standalone plugin with better configuration options. We fixed critical synchronization bugs where function responses could be corrupted by chart output, database flush deadlocks in the libuv pool, and idle CPU consumption on 32-bit time64 systems. SNMP and SMBIOS collectors are more robust, host labels stream properly through parents, and zstd-compressed archives reduce download sizes by 30%. Over 40 pull requests landed, mostly addressing race conditions, test flakiness, and protocol correctness.

New smbios_memory collector reports Linux physical memory inventory. Redfish gains BMC logs viewer and hardware enrichment. Legacy alarm notifications now map to Go providers. V2 charts restored.

This week's updates strengthen hardware monitoring and legacy system integration. A new smbios_memory collector provides Linux memory inventory tracking. The Redfish collector gains on-demand BMC logs viewing and expanded hardware details. Most significantly, legacy shell-format alarm notification configurations now map to 23 modern Go providers, enabling smoother migration paths for existing deployments.