Bounded-cardinality process grouping (macOS) and hardware sensors summaries, histogram and function - #23085
Merged
ktsaou merged 10 commits intoJul 11, 2026
Merged
Conversation
Copy link
Copy Markdown
Member
Sorry, something went wrong.
…rs summaries apps.plugin (macOS): - group processes by executable path: application bundles keep their own group; Apple framework helpers, standalone daemons and driver extensions aggregate into system-frameworks / system-daemons / driver-extensions (sealed-system-volume path classification; Apple bundle grammar is stable across macOS versions, so no per-version maintenance is needed) - executable path fallback chain: proc_pidpath() then KERN_PROCARGS2 - sshd-session added to the macOS process managers - stock apps_groups.conf: macOS components of interest (windowserver, spotlight, media-analysis, coreaudio, fseventsd, icloud, nsurlsessiond, timemachine, videotoolbox) - live result: ~600 process groups -> ~85 on an M4 Mac mini macos.plugin sensors: - fix: per-sensor chart label key renamed 'name' -> 'label' to match the Linux producers; same fix applied to windows.plugin GetSensors - fix: IOHID sensor identities no longer embed the per-boot IORegistryEntryID; unique product names are used as-is and duplicates disambiguate via LocationID, so identities are stable across reboots - fix: powermetrics loop mode used '-n 0', which means "zero samples and exit" on current macOS - thermal pressure and related charts were collected only briefly after each restart; also stop re-arming the log-once guard on every stored sample (it flooded the log with the same error every backoff cycle) - per-channel sensor charts are now opt-in for temperature, voltage, current and power; the default views are per-subsystem temperature min/avg/max and per-subsystem power totals (~405 sensor charts -> ~25 on an M4 Mac mini) hardware sensors, cross-OS: - new contract in src/collectors/common-contexts/hw-sensors.h: system.hw.sensor.temperature.histogram - a population histogram of all temperature sensors in fixed, disjoint buckets (40,50,60,70,80,85,90,95,100,+Inf upper bounds, degrees Celsius), identical on macOS, Windows and Linux (debugfs/libsensors), so Netdata Cloud can sum bucket counts across nodes for a fleet-wide thermal view - new 'sensors' function listing every discovered sensor with live readings: the per-sensor chart id is the table key (chart <-> row correlation), per-kind aggregation tiles, group-by schemes by kind/subsystem/device/state; registered only on hosts that actually have sensors; implemented on macOS and debugfs.plugin (the first function served by debugfs.plugin, via the functions event loop)
ktsaou
requested review from
Ancairon,
kanelatechnical and
thiagoftsm
as code owners
July 11, 2026 10:25
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 15 files
Confidence score: 5/5
- Safe to merge after the addressed issues were fixed.
sequenceDiagram
participant Kernel as macOS Kernel
participant apps as apps.plugin (macOS)
participant macosSensors as macos.plugin (sensors)
participant pmLoop as powermetrics loop
participant debugfs as debugfs.plugin (Linux)
participant winSensors as windows.plugin GetSensors
participant common as common-contexts/hw-sensors.h
participant funcEvloop as Functions Event Loop
participant RRDdb as Netdata Database
Note over apps,Kernel: CHANGED: Process Grouping (macOS)
apps->>Kernel: macos_get_executable_path(pid)
Kernel-->>apps: /System/Library/Frameworks/..., or /Applications/Xcode.app/...
apps->>apps: Parse path for .app, .framework, .dext, apple system prefix
alt Application bundle found (*.app, *.appex)
apps->>apps: Group by app name (e.g., "Finder")
else Apple framework helper
apps->>apps: Group => "system-frameworks"
else Apple standalone daemon (/usr/libexec, /usr/sbin, ...)
apps->>apps: Group => "system-daemons"
else Driver extension (*.dext)
apps->>apps: Group => "driver-extensions"
else Third-party framework
apps->>apps: Group by framework name
else Third-party plain binary (or launchd/interpreter)
apps->>apps: Fallback: group by comm/name (unchanged)
end
apps-->>RRDdb: Registered groups: ~600→~85 per host
Note over macosSensors,pmLoop: CHANGED: Sensors + powermetrics (macOS)
macosSensors->>macosSensors: NEW: do_per_sensor_charts[] (opt-in for temp/voltage/current/power)
macosSensors->>macosSensors: Collect all ~405 sensors (SMC + IOHID)
alt IOHID sensor identity
macosSensors->>macosSensors: CHANGED: Prefer LocationID over IORegistryEntryID
macosSensors->>macosSensors: CHANGED: label key replaces name in all chart labels
end
macosSensors->>macosSensors: Rollup per-subsystem (cpu/soc/mem/storage/ambient...)
Note over macosSensors: temperature_subsystem_* → min/avg/max
Note over macosSensors: power_subsystem_* → sum
macosSensors->>common: hw_sensors_temperature_histogram_add() per temp sensor
common-->>macosSensors: histogram bucket (0=≤40°C, 9=+Inf)
macosSensors-->>RRDdb: ~25 summary charts + histogram (default)
opt Per-sensor charts enabled
macosSensors-->>RRDdb: Per-channel charts also emitted (opt-in)
end
pmLoop->>pmLoop: CHANGED: Removed "-n 0" from powermetrics args
Note over pmLoop: -n 0 caused immediate exit streaming→dead loop
pmLoop->>pmLoop: Streams until netdata stops it
alt Productive cycle produced sample
pmLoop->>pmLoop: reset logged_loop_failure = false
pmLoop->>pmLoop: backoff = initial (100ms)
else No sample produced (backoff loop)
pmLoop->>pmLoop: exponential backoff, log-once guard NOT re-armed
end
pmLoop->>macosSensors: Store thermal/smc/gpu sample
macosSensors->>macosSensors: CHANGED: labels use "label" not "name"
Note over debugfs,RRDdb: NEW: Functions + Histogram (Linux debugfs.plugin)
debugfs->>funcEvloop: module_libsensors_register_functions(wg)
funcEvloop->>debugfs: function invocation (sensors table request)
debugfs->>debugfs: Lock sensors_data_mutex, iterate sensors_dict
debugfs-->>funcEvloop: JSON table: chart_id, label, kind, subsystem, device, reading, units, state, ...
debugfs->>debugfs: CHANGED: libsensors_emit_temperature_histogram()
Note over debugfs: Mirrors hw-sensors.h bucket contract via PLUGINSD text
debugfs-->>RRDdb: TEMP HISTOGRAM chart (sensors.temperature_histogram)
Note over winSensors,common: CHANGED: Windows GetSensors
winSensors->>common: hw_sensors_temperature_histogram_reset()
winSensors->>winSensors: Walk sensor dictionary
alt Sensor is temperature (NETDATA_WIN_SENSOR_CELSIUS)
winSensors->>common: hw_sensors_temperature_histogram_add(celsius)
end
winSensors->>common: common_hw_sensors_temperature_histogram(...)
common-->>RRDdb: TEMP HISTOGRAM chart (system.hw.sensor.temperature.histogram)
Note over winSensors: Label key fix: "label" not "name"
Note over RRDdb: Cross-OS histogram buckets (PERMANENT CONTRACT)
Note over RRDdb: Upper bounds: 40,50,60,70,80,85,90,95,100,+Inf °C
Reply with feedback, questions, or to request a fix.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/debugfs.plugin/debugfs_plugin.c
Comment thread
src/collectors/debugfs.plugin/module-libsensors.c
Comment thread
src/collectors/macos.plugin/macos_sensors.c
Comment thread
src/collectors/macos.plugin/macos_powermetrics.c
Comment thread
src/collectors/windows.plugin/GetSensors.c
Outdated
Comment thread
src/collectors/debugfs.plugin/debugfs_plugin.c
Comment thread
src/collectors/common-contexts/hw-sensors.h
…nction schema, sync metadata - debugfs: netdata_mutex_t is uv_mutex_t and has no static initializer - initialize sensors_data_mutex via a constructor (fixes the Linux builds) - debugfs: serialize ALL sensors_dict traversals (pre-flight, emission, histogram, function, cleanup) - the dictionary is single-threaded and is now accessed by both the collection thread and function workers - debugfs: the main loop honors the functions-evloop exit request (QUIT) and propagates its exit status; the function result is written under stdout_mutex so it cannot interleave with CHART/SET emission - debugfs: function rows use the cross-OS names (current/V/A/W, healthy state 'ok'), point state-only sensors at their _alarm chart, and report freshness from the current collection cycle; the histogram keeps emitting zero counts once exposed - macos: sensors cleanup runs under the sensors mutex (constructor initialized) and also frees the rollup list and obsoletes the rollup/histogram charts - windows: the histogram uses the exact chart-dimension value math, including the external-config multiplier override - share the 'sensors' function table schema via common-contexts/hw-sensors-function.h instead of duplicating it per producer; clarify the exclusive-upper-bound bucket semantics in hw-sensors.h; add both headers to the CMake sources list - metadata: rename the per-sensor label key 'name' to 'label' (macos, windows), document the new summary/histogram contexts and the new per-sensor chart config options
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 19 files
Confidence score: 5/5
- Safe to merge after the addressed issues were fixed.
sequenceDiagram
participant apps as apps.plugin (macOS)
participant os as macOS Kernel/Proc
participant macoS as macos.plugin
participant iokit as IOKit / IOHID
participant pm as powermetrics (ndsudo)
participant dbg as debugfs.plugin
participant libs as libsensors
participant win as windows.plugin (GetSensors)
participant cc as common-contexts (hw-sensors.h / hw-sensors-function.h)
participant nd as Netdata Daemon (rrdengine, functions)
Note over apps,os: PROCESS GROUPING (macOS)
loop per process
apps->>os: proc_pidpath(pid) or sysctl KERN_PROCARGS2
os-->>apps: executable path
apps->>apps: classify path (Apple vs user, .app/.framework/.dext)
alt matched by apps_groups.conf
apps->>apps: use configured group name
else unmatched
alt Apple framework helper
apps->>apps: group = "system-frameworks"
else Apple standalone daemon
apps->>apps: group = "system-daemons"
else driver extension
apps->>apps: group = "driver-extensions"
else user application bundle
apps->>apps: group = bundle basename
else user plain binary
apps->>apps: group = process name (comm)
end
end
apps->>nd: update chart dimensions for group
end
Note over macoS,pm: SENSOR COLLECTION & SUMMARY (macOS)
macoS->>iokit: scan SMC keys and IOHID services
macoS->>macOS: lock macos_sensors_mutex
Note over macoS: begin_cycle (mark all sensors unseen)
loop per sensor
macoS->>macoS: read sensor value (SMC/IOHID)
macoS->>macoS: store value, set seen, update last_value
alt per-sensor chart enabled for this kind
macoS->>nd: create/update per-sensor chart (rrdset)
else opt-in summary only
macoS->>macoS: skip per-sensor chart creation
end
end
macoS->>macoS: finish_cycle: accumulate histogram, rollups
macoS->>cc: hw_sensors_temperature_histogram_add() each temp sensor
macoS->>cc: common_hw_sensors_temperature_histogram() -> create/update heatmap chart
loop per subsystem
macoS->>macoS: create/update temperature.min/avg/max and power.sum rollup charts
end
macoS->>macoS: unlock mutex
Note over macoS,pm: POWERMETRICS LOOP FIX
pm->>pm: start powermetrics with -b 0 -i <ms> (removed -n 0)
pm->>macOS: stream plist samples
macoS->>pm: read from pipe
alt loop produced sample
macoS->>macoS: reset logged_loop_failure = false, backoff = initial
else no sample / failure
macoS->>macoS: backoff exponential backoff, keep logged_loop_failure true
end
macoS->>nd: store thermal pressure data
Note over dbg,libs: LINUX SENSORS (debugfs.plugin)
dbg->>nd: register "sensors" function via functions_evloop
loop collection cycle
dbg->>dbg: lock sensors_data_mutex
dbg->>libs: sensors_collect_data()
libs-->>dbg: sensor readings
dbg->>dbg: emit histogram via PLUGINSD text
alt sensors present
dbg->>nd: CHART sensors.temperature_histogram (heatmap)
dbg->>nd: SET each bucket count
end
dbg->>dbg: unlock mutex
end
Note over dbg: FUNCTION HANDLER
dbg->>nd: function request ('sensors')
dbg->>dbg: lock sensors_data_mutex
dbg->>libs: iterate sensors_dict
dbg->>dbg: build JSON row per sensor (chart_id, label, kind, ...)
dbg->>dbg: unlock mutex
dbg->>cc: hw_sensors_function_columns() + hw_sensors_function_presentation()
dbg->>nd: lock stdout_mutex, pluginsd_function_result_to_stdout()
Note over win,cc: WINDOWS SENSORS (histogram only)
win->>win: lock sensors_mutex
win->>win: iterate dictionary (collect sensors)
loop per sensor
win->>cc: hw_sensors_temperature_histogram_add() if type CELSIUS
end
win->>cc: common_hw_sensors_temperature_histogram() -> create/update heatmap
win->>win: unlock mutex
Note over cc,nd: CROSS-OS CONTRACTS
cc->>nd: define permanent histogram bucket edges (40,50,...100,+Inf)
cc->>nd: define sensors function table schema (columns, charts, grouping)
nd->>nd: Heatmap chart created only when temp sensors exist
Reply with feedback, questions, or to request a fix.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/debugfs.plugin/debugfs_plugin.c
…ion columns - debugfs: cancel and join the functions-evloop workers before module cleanup and the EXIT marker, on every exit path (early module-disable and EPIPE exits now break out of the loop instead of returning) - a worker running concurrently with shutdown could otherwise write a function result after EXIT or use freed state - hw-sensors-function.h: define the table columns as data and emit them in a loop instead of 17 repeated builder calls - macos: skip the IOHID pass entirely when no services exist; document that the duplicate-name scan is quadratic on a few dozen items at most - apps.plugin: document the intentionally asymmetric application/framework precedence in the bundle parser
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 19 files
Confidence score: 5/5
- Safe to merge after the addressed issues were fixed.
sequenceDiagram
participant Client as Netdata Client/UI
participant Agent as Netdata Agent
participant AppsPlugin as apps.plugin
participant MacOSPlugin as macos.plugin (sensors)
participant LinuxPlugin as debugfs.plugin (libsensors)
participant WindowsPlugin as windows.plugin (GetSensors)
participant Kernel as macOS/IOKit/SMC
participant Histogram as Temperature Histogram "system.hw.sensor.temperature.histogram"
participant Functions as "sensors" Function
participant Config as apps_groups.conf
Note over Agent,Functions: NEW: Cross-OS Sensor Features
Agent->>Histogram: NEW: Publish temperature histogram across all OSes
Note over Histogram: Fixed buckets: 40,50,60,70,80,85,90,95,100,+Inf °C
Agent->>Functions: NEW: Register "sensors" function (only on hosts with sensors)
rect rgb(200, 200, 255)
Note over MacOSPlugin: macOS Sensor Changes
MacOSPlugin->>MacOSPlugin: CHANGED: Use "label" instead of "name" label key
MacOSPlugin->>MacOSPlugin: FIXED: Remove IORegistryEntryID from sensor identity
MacOSPlugin->>MacOSPlugin: Use stable LocationID for duplication
MacOSPlugin->>Kernel: FIXED: Remove -n 0 from powermetrics loop
Kernel-->>MacOSPlugin: Stream data continuously (no early exit)
MacOSPlugin->>MacOSPlugin: CHANGED: Per-channel charts are opt-in for temperature/voltage/current/power
end
rect rgb(255, 200, 200)
Note over AppsPlugin: macOS Process Grouping
AppsPlugin->>Config: Read apps_groups.conf (with new macOS entries)
AppsPlugin->>AppsPlugin: NEW: Path-based grouping for unmatched processes
alt Application bundles (.app, .appex)
AppsPlugin->>AppsPlugin: Group per application (e.g., Finder, Xcode)
else Apple framework helpers (.framework)
AppsPlugin->>AppsPlugin: Aggregate into "system-frameworks"
else Apple standalone daemons (/usr/libexec, /usr/sbin, ...)
AppsPlugin->>AppsPlugin: Aggregate into "system-daemons"
else Driver extensions (.dext)
AppsPlugin->>AppsPlugin: Aggregate into "driver-extensions"
else Third-party frameworks
AppsPlugin->>AppsPlugin: Group per framework
else Third-party plain binaries
AppsPlugin->>AppsPlugin: Group per process name (comm)
end
end
par macOS
MacOSPlugin->>MacOSPlugin: NEW: Build per-subsystem temperature summary (min/avg/max)
MacOSPlugin->>MacOSPlugin: NEW: Build per-subsystem power summary (sum)
MacOSPlugin->>Histogram: NEW: Feed temperature histogram
MacOSPlugin->>Functions: NEW: List all sensors in response
and Linux
LinuxPlugin->>LinuxPlugin: CHANGED: Serialize sensors_dict access with mutex
LinuxPlugin->>LinuxPlugin: CHANGED: Write function results under stdout_mutex
LinuxPlugin->>LinuxPlugin: CHANGED: Cancel/join function threads on exit
LinuxPlugin->>Histogram: NEW: Emit temperature histogram via PLUGINSD text
LinuxPlugin->>Functions: NEW: First function (using functions_event_loop)
and Windows
WindowsPlugin->>WindowsPlugin: CHANGED: Use "label" instead of "name" label key
WindowsPlugin->>WindowsPlugin: CHANGED: Include external multiplier in histogram math
WindowsPlugin->>Histogram: NEW: Feed temperature histogram
WindowsPlugin->>Functions: NEW: (Pending on this branch)
end
Agent-->>Client: Serve summary charts (25 instead of 405)
Agent-->>Client: Serve process groups (85 instead of 598)
Agent-->>Client: Serve temperature histogram
Agent-->>Client: Serve "sensors" function table
Note over Agent,Client: Result: Mac M4 Mini → 598→85 process groups, 405→25 sensor charts
Reply with feedback, questions, or to request a fix.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/macos.plugin/macos_sensors.c
- debugfs: the histogram chart-exposed flag moves to file scope and is reset at cleanup, so a future in-process module restart re-emits the CHART definition instead of sending orphan SETs - debugfs: sensor states are refreshed in the same locked section as the collection, so a concurrent 'sensors' function request never observes values derived from the previous cycle - windows: guard the histogram insertion with isfinite(), matching the validation the macOS and libsensors producers already apply
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 20 files
Confidence score: 5/5
- Safe to merge after the addressed issues were fixed.
sequenceDiagram
participant AppsPlugin as apps.plugin (macOS)
participant MacProcess as macOS Process
participant MacSensorPlugin as macos.plugin sensors
participant DebugfsPlugin as debugfs.plugin libsensors
participant WinPlugin as windows.plugin GetSensors
participant IOHID as IOHID / SMC
participant Powermetrics as powermetrics
participant Histogram as HW Sensors Histogram
participant SensorFunc as Sensors Function
participant NetdataCloud as Netdata Cloud / UI
Note over AppsPlugin: Path-Based Process Grouping (macOS)
MacProcess->>AppsPlugin: PID, executable path (/System, .app, .framework, .dext)
AppsPlugin->>AppsPlugin: NEW: Classify path (apple vs user, bundle type)
alt Application bundle (*.app / *.appex)
AppsPlugin->>AppsPlugin: Group by app name (e.g. Finder)
else Apple framework helper (*.framework, /System)
AppsPlugin->>AppsPlugin: Group "system-frameworks"
else Apple standalone daemon (/usr/libexec, /usr/sbin, /bin)
AppsPlugin->>AppsPlugin: Group "system-daemons"
else Driver extension (*.dext)
AppsPlugin->>AppsPlugin: Group "driver-extensions"
else Third-party framework
AppsPlugin->>AppsPlugin: Group by framework name
else Third-party plain binary
AppsPlugin->>AppsPlugin: Fallback to process comm name
end
Note over AppsPlugin: Result: ~600 groups -> ~85 groups
Note over MacSensorPlugin,IOHID: Sensor Identity Stability & Label Fix
IOHID->>MacSensorPlugin: Sensor with IORegistryEntryID (per-boot)
MacSensorPlugin->>MacSensorPlugin: NEW: Use LocationID for stable identity
MacSensorPlugin->>MacSensorPlugin: Use unique product names (e.g. PMU_tdie5)
MacSensorPlugin->>MacSensorPlugin: CHANGED: Label key "name" -> "label"
Note over MacSensorPlugin,Powermetrics: powermetrics Loop Fix
Powermetrics->>MacSensorPlugin: -n 0 (zero samples, exit immediately)
MacSensorPlugin->>MacSensorPlugin: CHANGED: Remove -n flag, stream continuously
MacSensorPlugin->>MacSensorPlugin: NEW: Reset log-once guard only after productive loop
Note over MacSensorPlugin,Histogram: Sensor Cardinality Reduction
MacSensorPlugin->>MacSensorPlugin: NEW: Per-channel charts opt-in (default off for temp/volt/current/power)
MacSensorPlugin->>Histogram: Build per-subsystem temperature summary (min/avg/max)
MacSensorPlugin->>Histogram: Build per-subsystem power summary (total watts)
MacSensorPlugin->>Histogram: Build temperature histogram (10 disjoint buckets)
Note over DebugfsPlugin: Linux libsensors Integration
DebugfsPlugin->>DebugfsPlugin: NEW: Register "sensors" function via functions event loop
DebugfsPlugin->>Histogram: Mirror histogram contract (PLUGINSD text protocol)
DebugfsPlugin->>DebugfsPlugin: NEW: sychronized sensor data access for function threads
DebugfsPlugin->>DebugfsPlugin: NEW: Graceful shutdown handling (QUIT, EPIPE)
Note over WinPlugin: Windows Sensor Integration
WinPlugin->>WinPlugin: CHANGED: Label key "name" -> "label"
WinPlugin->>Histogram: Count temperatures only on successful/finite reads
WinPlugin->>Histogram: Match histogram value math with per-sensor charts
WinPlugin->>WinPlugin: NEW: Retire histogram chart on cleanup
Note over Histogram: Cross-OS Temperature Histogram Contract
Histogram->>NetdataCloud: Fixed bucket edges (40,50,60,70,80,85,90,95,100,+Inf °C)
Histogram->>NetdataCloud: Fleet-aggregatable (bucket counts sum across nodes)
Note over Histogram: Permanent contract - never change bucket definitions
Note over SensorFunc: Cross-OS Sensors Function
SensorFunc->>NetdataCloud: One row per sensor
SensorFunc->>NetdataCloud: Columns: chart_id, label, kind, subsystem, source, device, sensor_id, reading, units, state, chart_mode
SensorFunc->>NetdataCloud: Per-kind columns (Temperature/Fan/Voltage/Current/Power)
SensorFunc->>NetdataCloud: Groupable by kind, subsystem, device, state
SensorFunc->>NetdataCloud: Registration follows discovery (only on hosts with sensors)
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/windows.plugin/GetSensors.c
Sensors absent from the Sensor API enumeration kept their last read_ok and cached value, so the temperature histogram counted them forever. Every enumerated sensor is now stamped with the enumeration generation, which advances only after a completed pass: a sensor missing from one completed pass stops being counted, sensors not yet visited by the in-flight pass keep one pass of grace, and a failed poll freezes the last known state instead of aging everything out.
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
No issues found across 20 files
Confidence score: 5/5
- Automated review surfaced no issues in the provided summaries.
- No files require special attention.
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/macos.plugin/macos_sensors.c
Interpreters already fall back to comm naming in the macOS tree-target hook, but the default interpreters list was missing zsh - the macOS default login shell - and the other login shells macOS ships. An SSH session (a shell that is a tree root under the sshd-session manager) was therefore bucketed into system-daemons together with everything the user ran in it. Add zsh, dash, csh, tcsh and ksh to the defaults, so shell tree roots group by their comm on all platforms alike.
The libsensors module had no metadata entry at all while this branch adds a new public context and a function to it. Document the module: per-sensor input/average/alarm contexts for every sensor class (only the average contexts libsensors actually provides - voltage, current, power), the chart labels, the alarm states, and the cross-OS temperature histogram with its fixed bands. Also document why macos_sensors_mutex deliberately has no destructor: the daemon exits with collector threads alive, and destroying a held uv mutex aborts.
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
All reported issues were addressed across 21 files
Confidence score: 5/5
- Safe to merge after the addressed issues were fixed.
sequenceDiagram
participant MacOS as macOS System
participant AppsPlugin as apps.plugin
participant Targets as Target Tree
participant MacSensors as macos.plugin (sensors)
participant WinSensors as windows.plugin (GetSensors)
participant DebugFS as debugfs.plugin (libsensors)
participant Histogram as Temperature Histogram
participant FuncAPI as Sensors Function API
participant RRD as Netdata DB / Charts
participant Powermetrics as powermetrics
Note over MacOS,Targets: NEW: Path-derived process grouping (macOS)
MacOS->>AppsPlugin: collect process list
AppsPlugin->>AppsPlugin: get_executable_path(pid)
alt Executable path available
AppsPlugin->>AppsPlugin: classify by path (app, framework, dext, system, user)
alt Application bundle (*.app, *.appex)
AppsPlugin->>Targets: group = app name
else Apple framework helper
AppsPlugin->>Targets: group = "system-frameworks"
else Apple standalone daemon (/usr/libexec, /usr/sbin, ...)
AppsPlugin->>Targets: group = "system-daemons"
else Driver extension (*.dext)
AppsPlugin->>Targets: group = "driver-extensions"
else Third-party framework
AppsPlugin->>Targets: group = framework name
else Third-party binary
AppsPlugin->>Targets: fallback to comm naming
end
else No executable path
AppsPlugin->>Targets: fallback to comm naming
end
AppsPlugin->>Targets: add to target tree
Note over MacOS,Targets: Result: ~598 groups → ~85 on M4 Mac mini
Note over MacOS,Powermetrics: CHANGED: powermetrics loop fix
Powermetrics->>Powermetrics: removed -n 0 (streams until stopped)
Powermetrics->>MacOS: continuous data stream
MacSensors->>Powermetrics: CHANGED: reset logged_loop_failure only after productive loop
Powermetrics-->>MacSensors: thermal/smc/gpu data
Note over MacSensors,Histogram: NEW: sensor summaries + histogram
MacSensors->>MacSensors: collect all sensor values
MacSensors->>MacSensors: only chart per-channel for do_per_sensor_charts[kind]=true
alt Temperature sensor
MacSensors->>Histogram: hw_sensors_temperature_histogram_add(value)
MacSensors->>MacSensors: update temperature_subsystem_* rollup (min/avg/max)
else Power sensor
MacSensors->>MacSensors: update power_subsystem_* rollup (sum)
end
MacSensors->>RRD: create/update system.hw.sensor.temperature.histogram
MacSensors->>RRD: create/update system.hw.sensor.temperature.subsystem.*
MacSensors->>RRD: create/update system.hw.sensor.power.subsystem.*
Note over WinSensors,Histogram: CHANGED: Windows histogram integration
WinSensors->>WinSensors: collect sensor data
WinSensors->>WinSensors: CHANGED: use 'label' key instead of 'name'
WinSensors->>WinSensors: track read_ok per sensor
WinSensors->>WinSensors: track seen_generation vs sensors_enum_generation
alt Temperature sensor with successful read
WinSensors->>Histogram: hw_sensors_temperature_histogram_add(celsius)
end
WinSensors->>RRD: create/update system.hw.sensor.temperature.histogram
Note over DebugFS,Histogram: NEW: libsensors histogram + functions
DebugFS->>DebugFS: read hwmon sensors via libsensors
DebugFS->>Histogram: emit PLUGINSD CHART/DIMENSION/BEGIN/SET
DebugFS->>RRD: system.hw.sensor.temperature.histogram (heatmap)
DebugFS->>FuncAPI: register "sensors" function
Note over FuncAPI,RRD: NEW: cross-OS sensors function
Client->>FuncAPI: request sensors function
FuncAPI->>MacSensors: read sensor data (under mutex)
FuncAPI->>DebugFS: read sensor data (under mutex)
FuncAPI->>FuncAPI: build table rows with per-kind columns
FuncAPI-->>Client: JSON table (chart, label, kind, subsystem, source, device, reading, units, state, per-kind values)
Note over FuncAPI,Client: Same schema on macOS, Linux, Windows
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Re-trigger cubic
Sorry, something went wrong.
Comment thread
src/collectors/debugfs.plugin/metadata.yaml
Outdated
- windows.plugin: destroy the sensors mutex after the dictionary, so a future dictionary delete callback can never observe a destroyed mutex - macos.plugin: document why the function-registration check reads the sensors list without the mutex (collection, the check, and the pthread cleanup handler all run on the plugin thread; function workers lock) - debugfs.plugin metadata: fix the Additional typo in all module entries
The collector taxonomy checker requires a sibling taxonomy.yaml when a collector's metadata metrics change. Follow the macos.plugin and windows.plugin precedent: taxonomy_optout with a reason, since dashboard placement for these heterogeneous kernel-subsystem contexts is not yet defined.
Copy link
Copy Markdown
Member
Author
|
@cubic-dev-ai please review again |
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
@ktsaou I have started the AI code review. It will take a few minutes to complete. |
Sorry, something went wrong.
Copy link
Copy Markdown
|
Sorry, something went wrong.
Copy link
Copy Markdown
Contributor
There was a problem hiding this comment.
No issues found across 22 files
Confidence score: 5/5
- Automated review surfaced no issues in the provided summaries.
- No files require special attention.
You've manually re-run cubic several times on this PR. Each manual re-review checks the full PR again and counts toward your usage quota. To preserve your usage limits, we recommend letting cubic automatically review new commits.
Re-trigger cubic
Sorry, something went wrong.
stelfrag
added a commit
to stelfrag/netdata
that referenced
this pull request
Jul 12, 2026
…#23085) Backport the macos.plugin sensor collectors (gpu, sensors/SMC, power, powermetrics, nvme, iohid) and the shared hw-sensors histogram/function framework to the 2.10.x line. macOS-only paths taken from master tip; the functions_evloop/logs/network-viewer/status-file changes bundled in those PRs are excluded (not needed by the in-process macOS collector). Adds the ndsudo powermetrics whitelist (with <errno.h>) and CMake wiring.
Merged
stelfrag
pushed a commit
to stelfrag/netdata
that referenced
this pull request
Jul 14, 2026
Backport the apps.plugin half of netdata#23085, the companion to the macOS sensor collectors already backported. Adds apps_os_tree_target_name_macos() (derive process groups from the executable path so macOS does not create one group per distinct process name) plus the macOS component definitions in apps_groups.conf. apps_os_tree_target_name() is a compile-time no-op on non-macOS (returns NULL -> existing name/comm naming), so Linux/Windows/ FreeBSD behavior is unchanged. Depends only on the managed_list/tree mechanism already present since netdata#18594. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ferroin
pushed a commit
that referenced
this pull request
Jul 15, 2026
Backport the macos.plugin sensor collectors (gpu, sensors/SMC, power, powermetrics, nvme, iohid) and the shared hw-sensors histogram/function framework to the 2.10.x line. macOS-only paths taken from master tip; the functions_evloop/logs/network-viewer/status-file changes bundled in those PRs are excluded (not needed by the in-process macOS collector). Adds the ndsudo powermetrics whitelist (with <errno.h>) and CMake wiring.
Ferroin
pushed a commit
that referenced
this pull request
Jul 15, 2026
Backport the apps.plugin half of #23085, the companion to the macOS sensor collectors already backported. Adds apps_os_tree_target_name_macos() (derive process groups from the executable path so macOS does not create one group per distinct process name) plus the macOS component definitions in apps_groups.conf. apps_os_tree_target_name() is a compile-time no-op on non-macOS (returns NULL -> existing name/comm naming), so Linux/Windows/ FreeBSD behavior is unchanged. Depends only on the managed_list/tree mechanism already present since #18594. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
1 task
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Reviewers
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
This PR fixes runaway per-node cardinality on macOS (processes and hardware sensors), fixes three sensor bugs, and introduces two cross-OS conventions: a fleet-aggregatable temperature histogram and a sensors function.
apps.plugin: path-derived process grouping (macOS)
macOS has a flat process tree (launchd spawns almost everything) and the stock apps_groups.conf is Linux-oriented, so every unmatched process became its own group: ~600 groups (~6,000 time-series) per Mac.
Processes not matched by apps_groups.conf are now grouped by their executable path:
Apple's bundle grammar and the sealed system volume (SIP/SSV) make this classification stable across macOS versions - no name lists to maintain. Famous components (WindowServer, Spotlight, media analysis, CoreAudio, iCloud, Time Machine, VideoToolbox, ...) are re-exposed as named groups via the stock apps_groups.conf, and any aggregated component can be re-exposed with one config line.
Live result on an M4 Mac mini: 598 groups → 85, with system-frameworks/system-daemons absorbing ~85% of all processes while contributing almost no CPU to the buckets (WindowServer, previously hidden noise among 600 groups, is now clearly visible as the top consumer).
macos.plugin sensors: three fixes
sensors cardinality (macOS)
An M4 Mac mini exposes ~405 sensors (202 temperatures, ~150 per-regulator voltage/current rails). Per-channel charts are now opt-in for temperature/voltage/current/power; the default views are:
Live result: ~405 sensor charts → ~25, everything still collected internally, per-channel detail one config switch away.
Cross-OS temperature histogram
New permanent contract (src/collectors/common-contexts/hw-sensors.h): system.hw.sensor.temperature.histogram - a population histogram of ALL temperature sensors in fixed, disjoint buckets (upper bounds 40,50,60,70,80,85,90,95,100,+Inf °C; 10°C bands in the benign range, 5°C bands in the 80-100°C action zone). Implemented identically on macOS, Windows (GetSensors) and Linux (debugfs.plugin/libsensors), so bucket counts can be summed across nodes for a fleet-wide thermal distribution view. The chart is only created on hosts that have temperature sensors.
sensors function
A live table of every discovered sensor - including the ones aggregated into summaries: label, kind, subsystem, source, device, current reading, units, state. The per-sensor chart id is the table's unique key (chart ↔ row correlation), per-kind aggregation tiles (Temperature/Fan/Voltage/Current/Power) re-aggregate under user-selected groupings (by kind, subsystem, device, state). Registered only on hosts that actually have sensors.
Implemented on macOS and debugfs.plugin (its first function, served via the functions event loop). The Windows implementation follows on this branch.
Validation
Notes for reviewers
Summary by cubic
Caps macOS process and sensor cardinality and adds fleet‑ready sensor views across OSes. Introduces path‑based process grouping, per‑subsystem summaries, a cross‑OS temperature histogram, and a unified sensors function, with fixes for labels, identity, stream loops, and hardened histogram, shutdown, and data validity paths.
New Features
Bug Fixes
Written for commit 99bc2af. Summary will update on new commits.