Host Metrics

The Host Metrics source collects system metrics from the host the collector runs on, using the OpenTelemetry hostmetricsreceiver scrapers. It gathers CPU, memory, disk, filesystem, network, paging, load, and process-count metrics, and can optionally collect per-process metrics. Choose which metric categories and individual metrics to scrape, and how often.

Supported Telemetry Types

Platform	Metrics	Logs	Traces
Linux	✓
Windows	✓
macOS	✓
AIX	✓

Prerequisites

The collector must run directly on the host you want to monitor. It needs read access to the host's system interfaces:

• On Linux, read access to /proc and /sys. Per-process metrics require running the collector as root.

• On Windows, per-process metrics require running the collector as Administrator.

• On macOS and AIX, the collector reads the available system interfaces. Per-process metrics are not supported on AIX.

Metric and scraper availability varies by operating system. See the upstream Host Metrics Receiver documentation for metric-level OS compatibility.

Configuration

Metrics

The Metrics parameter selects which metric categories and individual metrics the source collects. Each category below can be expanded in the UI to toggle its metrics individually. The Default column reflects the upstream Host Metrics Receiver (v0.153.0); metric and scraper availability still varies by operating system (see Prerequisites).

Filesystem Metrics

Metric	Unit	Default	Description
system.filesystem.inodes.usage	{inodes}	Enabled	FileSystem inodes used.
system.filesystem.usage	By	Enabled	Filesystem bytes used.
system.filesystem.utilization	1	Disabled	Fraction of filesystem bytes used.

Memory Metrics

Metric	Unit	Default	Description
system.memory.usage	By	Enabled	Bytes of memory in use.
system.memory.utilization	1	Disabled	Percentage of memory bytes in use.

Network Metrics

Metric	Unit	Default	Description
system.network.connections	{connections}	Enabled	The number of connections.
system.network.conntrack.count	{entries}	Disabled	The count of entries in conntrack table.
system.network.conntrack.max	{entries}	Disabled	The limit for entries in the conntrack table.
system.network.dropped	{packets}	Enabled	The number of packets dropped.
system.network.errors	{errors}	Enabled	The number of errors encountered.
system.network.io	By	Enabled	The number of bytes transmitted and received.
system.network.packets	{packets}	Enabled	The number of packets transferred.

Paging Metrics

Metric	Unit	Default	Description
system.paging.faults	{faults}	Enabled	The number of page faults.
system.paging.operations	{operations}	Enabled	The number of paging operations.
system.paging.usage	By	Enabled	Swap (unix) or pagefile (windows) usage.
system.paging.utilization	1	Disabled	Swap (unix) or pagefile (windows) utilization.

Load Metrics

Metric	Unit	Default	Description
system.cpu.load_average.1m	{thread}	Enabled	Average CPU Load over 1 minute.
system.cpu.load_average.5m	{thread}	Enabled	Average CPU Load over 5 minutes.
system.cpu.load_average.15m	{thread}	Enabled	Average CPU Load over 15 minutes.

CPU Metrics

Metric	Unit	Default	Description
system.cpu.time	s	Enabled	Total seconds each logical CPU spent on each mode.
system.cpu.utilization	1	Disabled	Difference in system.cpu.time since the last measurement per logical CPU, divided by the elapsed time (value in interval [0,1]).

Disk Metrics

Metric	Unit	Default	Description
system.disk.io	By	Enabled	Disk bytes transferred.
system.disk.io_time	s	Enabled	Time disk spent activated. On Windows, this is calculated as the inverse of disk idle time.
system.disk.merged	{operations}	Enabled	The number of disk reads/writes merged into single physical disk access operations.
system.disk.operation_time	s	Enabled	Time spent in disk operations.
system.disk.operations	{operations}	Enabled	Disk operations count.
system.disk.pending_operations	{operations}	Enabled	The queue size of pending I/O operations.
system.disk.weighted_io_time	s	Enabled	Time disk spent activated multiplied by the queue length.

Processes Metrics

Metric	Unit	Default	Description
system.processes.count	{processes}	Enabled	Total number of processes in each state.
system.processes.created	{processes}	Enabled	Total number of created processes. Supported on Linux and OpenBSD.

System Metrics

Metric	Unit	Default	Description
system.uptime	s	Enabled	The time the system has been running.

Process Metrics

Parameter	Type	Required	Default	Description
Process Metrics	Boolean	No	true	Enable to collect per-process metrics. Compatible with Linux and Windows. The collector must run as root (Linux) and Administrator (Windows). Not supported on AIX.
Mute Process Name Errors	Boolean	No	true	Enable to prevent process name errors from being logged. Relevant when Process Metrics is true.
Mute Process EXE Errors	Boolean	No	true	Enable to prevent process exe lookup errors from being logged. Relevant when Process Metrics is true.
Mute Process IO Errors	Boolean	No	false	Enable to prevent input output errors from being logged. Relevant when Process Metrics is true.
Mute Process Username Errors	Boolean	No	false	Enable to prevent process username lookup errors from being logged. Relevant when Process Metrics is true.
Process Name Filtering	Boolean	No	false	Enable to configure include/exclude filtering for process metrics. Relevant when Process Metrics is true.
Process Include Filter	Strings	No	(empty)	List of processes to include for metric collection. Defaults to all processes. Relevant when Process Name Filtering is true.
Process Exclude Filter	Strings	No	(empty)	List of processes to exclude from metric collection. Relevant when Process Name Filtering is true.
Process Filter Match Type	Enum: regexp, strict	No	regexp	Strategy for matching process names against the include and exclude filters. Relevant when Process Name Filtering is true.

When Process Metrics is enabled, the following per-process metrics can be toggled individually.

Process Metric Toggles

Metric	Unit	Default	Description
process.context_switches	{count}	Disabled	Number of times the process has been context switched.
process.cpu.time	s	Enabled	Total CPU seconds broken down by different states.
process.cpu.utilization	1	Disabled	Percentage of total CPU time used by the process since last scrape, expressed as a value between 0 and 1. On the first scrape, no data point is emitted for this metric.
process.disk.io	By	Enabled	Disk bytes transferred.
process.disk.operations	{operations}	Disabled	Number of disk operations performed by the process.
process.memory.usage	By	Enabled	The amount of physical memory in use.
process.memory.utilization	1	Disabled	Percentage of total physical memory that is used by the process.
process.memory.virtual	By	Enabled	Virtual memory size.
process.open_file_descriptors	{count}	Disabled	Number of file descriptors in use by the process.
process.paging.faults	{faults}	Disabled	Number of page faults the process has made.
process.signals_pending	{signals}	Disabled	Number of pending signals for the process.
process.threads	{threads}	Disabled	Process threads count.
process.handles	{count}	Disabled	Number of open handles held by the process.

Advanced

Parameter	Type	Required	Default	Description
Collection Interval	Integer	No	60	How often (seconds) to scrape for metrics.

Examples

Collect host metrics with per-process metrics for a named set of processes

This configuration collects the default metric categories every 30 seconds and gathers per-process metrics only for processes whose names match nginx or postgres.

apiVersion: bindplane.observiq.com/v1
kind: Source
metadata:
  name: host
spec:
  type: host
  parameters:
    - name: collection_interval
      value: 30
    - name: enable_process
      value: true
    - name: enable_process_filter
      value: true
    - name: process_include
      value:
        - nginx
        - postgres
    - name: process_filter_match_strategy
      value: regexp

Configuration Tips

• Per-process metrics require elevated privileges. Run the collector as root on Linux or as Administrator on Windows, or the process scraper will fail to read process details.

• If process name, exe, IO, or username lookups produce repeated errors in the collector log, enable the corresponding Mute options to suppress the noise.

• Lower the Collection Interval for finer resolution at the cost of more data volume. The default is 60 seconds.

Troubleshooting

No process metrics are collected

Symptoms: per-process metrics are missing even though Process Metrics is enabled.

Solutions:

1. Confirm the collector is running as root (Linux) or Administrator (Windows). Per-process metrics require elevated privileges.

2. Verify the host is not AIX. Per-process metrics are not supported on AIX.

3. If Process Name Filtering is enabled, check that the include and exclude filters and the match type actually match the running process names.

Repeated process lookup errors in the collector log

Symptoms: the collector log fills with process name, exe, IO, or username lookup errors.

Solutions:

1. Enable the relevant Mute option (Mute Process Name Errors, Mute Process EXE Errors, Mute Process IO Errors, or Mute Process Username Errors).

2. These errors are common when the collector cannot read details for short-lived or privileged processes.

A metric category produces no data on a given OS

Symptoms: an enabled scraper returns no metrics on a particular platform.

Solutions:

1. Metric and scraper availability varies by operating system. Check the Host Metrics Receiver documentation for metric-level OS compatibility.

Standalone Source

apiVersion: bindplane.observiq.com/v1
kind: Source
metadata:
  name: host
spec:
  type: host
  parameters:
    - name: collection_interval
      value: 60
    - name: enable_process
      value: true
    - name: mute_process_name_error
      value: true
    - name: mute_process_exe_error
      value: true
    - name: mute_process_io_error
      value: false
    - name: mute_process_user_error
      value: false
    - name: enable_process_filter
      value: false

Related Resources

• Host Metrics Receiver — the upstream OpenTelemetry receiver this source configures.

• Resource Detection Processor — added automatically to attach host resource attributes.