Metrics Export

The OPC Router continuously collects runtime metrics and can export them to external monitoring systems such as Grafana, Prometheus, or InfluxDB. A subset of the metrics can be viewed directly in the Service Dashboard.

Enterprise+ add-on license required

Exporting metrics via OTLP or InfluxDB requires an active Enterprise+ license. For more information on license assignment and other Enterprise+ features, see License Levels. Some system-wide metrics (CPU, memory, drives) are recorded internally regardless of this and displayed in the Service Dashboard.

Configuration

You can find the settings in the menu under Options → Monitoring and Diagnostics, in the Telemetry and Log Export Settings section. This same page also consolidates the log output and internal metric logging; the cards Zentrale Metriksammlung über OTLP and Zentrale Metriksammlung über InfluxDB are responsible for the export. The “Zentrale Log-Erfassung über OTLP” dashboard exports logs instead of metrics and is not described here.

OTLP

Property	Description
Enable OTLP Metrics Collection	Enables OTLP export
OTLP Endpoint	Destination address of the OTLP collector, e.g., http://localhost:4317
OTLP Protocol	Transport protocol: gRPC (default) or HTTP/Protobuf
OTLP Headers	Optional HTTP headers, e.g., for authentication (Authorization: Bearer ...)
Timeout (ms)	Time limit for transmission to the collector (default: 10,000 ms)

InfluxDB

Property	Description
Enable metric export to InfluxDB	Enables InfluxDB export
InfluxDB endpoint	URL of the InfluxDB instance, e.g., http://localhost:8086
InfluxDB Organization	InfluxDB organization
InfluxDB Bucket	Target bucket
InfluxDB Token	API token for authentication
Flush Interval (ms)	How often collected metrics are written (default: 1,000 ms)
Export Interval (ms)	Measurement Interval (default: 60,000 ms)

note

The InfluxDB token and the OTLP headers are stored directly in the fields; they cannot be selected from the central secrets repository. Therefore, assign a token that has only write permissions for the bucket being used.

System Instrumentation

The switches are located on the same page in the map at Zusätzliche Erfassung von Metriken.

Property	Description	Default
Enable runtime metrics	Runtime metrics for the .NET runtime (GC, thread pool, ...)	Enabled
Enable process metrics	Process metrics (CPU time, memory)	Enabled
Enable ASP.NET Core metrics	HTTP server metrics	Disabled
Enable HTTP client metrics	Metrics for outgoing HTTP requests	Disabled

---

How to Read the Metrics

When analyzing the metrics, not only the metric name but also the metric type is important:

Type	Meaning	Typical Use
Gauge	Snapshot of a current state	Current load, queue size, status
Counter	Monotonically increasing counter	Number of errors, successful executions, connection attempts
Histogram	Distribution of measured values	Latences, processing times, message sizes

Many metrics are also exported with attributes, such as the flow name. This allows the same metric to be analyzed separately for each flow, plug-in, or target system.

Combine multiple metrics for a more meaningful analysis

Individual metrics are often only meaningful when combined. For example, an increasing queue depth is only problematic if worker utilization remains high or processing times increase at the same time.

Cardinality: Time Series per Metric

Each attribute combination of a metric—also known as a tag combination in monitoring systems—forms its own time series. Storage is reserved for these time series as soon as metric collection begins—even without an active export, since the service dashboard uses the same collection.

By default, a maximum of 128 time series are tracked individually for each metric. The collection aggregates all additional attribute combinations into a single overflow point. The values in this overflow point can no longer be analyzed by attribute.

If a system—for example, due to a very large number of flows, triggers, or plug-in instances—generates more than 128 attribute combinations for a metric, increase the limit using the environment variable OR_METRICS_CARDINALITY_LIMIT. A higher limit increases the memory requirements per service process. The change takes effect after the services are restarted. See Environment Variables for Metrics.

Value mapping for status metrics

Some status metrics are exported as numeric values. Therefore, a fixed value table should be used for dashboards and alerts.

inray.service.escalation_level

Describes the global health status of the OPC router.

Value	Meaning	Practical Interpretation
-2	Undefined	Status cannot yet be determined
-1	Initializing	Service is still initializing internal components
0	EscalationLevel0	Normal operation
1	EscalationLevel1	Good, highly repetitive transfer values are partially discarded
2	EscalationLevel2	Good transfer values are discarded
3	EscalationLevel3	Transfer values are completely discarded
4	EscalationLevel4	Messages with normal priority are discarded

inray.connections.state

Describes the current state of a flow.

Value	Meaning	Practical Interpretation
-1	Null	No valid flow state yet
0	StartUp	Flow is starting
1	Validate	Flow is being validated
2	Ready	Flow is ready for operation
3	TriggerCheckError	Error during trigger check
4	TriggerStatusError	Error in trigger status
5	TriggerConfigError	Error in trigger configuration
6	PlugInError	Error in a participating plug-in
7	NoLicense	License is missing or invalid
8	ShutDown	Flow is being terminated or has been shut down
9	StandBy	Flow is inactive but exists

inray.connections.execution.state

Describes the result of the last executed flow.

Value	Meaning	Practical Interpretation
0	None	No execution yet or no result available
1	Ok	Last execution successful
2	Aborted	Execution was aborted
3	Error	Execution ended with an error
4	Pending	Execution is still pending
5	Forwarded	Execution was forwarded
6	Discarded	Execution was discarded

inray.connections.trigger.state and inray.connections.trigger.async.status

Describes the state of a flow’s trigger.

Value	Meaning	Practical Interpretation
0	None	No valid trigger status available yet
1	Ok	Trigger is functioning properly
2	StatusError	Trigger reports a status error
3	CheckError	Trigger check failed
4	ConfigError	Trigger is configured incorrectly

inray.plugins.instance.status

Describes the status of a plugin instance.

Value	Meaning	Practical Interpretation
0	Uninitialized	The plugin has not yet been initialized
1	Initializing	The plugin is starting up or initializing resources
2	Ok	The plugin is working properly
3	Warning	The plug-in is running but is reporting warnings
4	Error	The plug-in is malfunctioning
5	Stopped	The plug-in has been stopped
6	NoLicense	Plug-in cannot run due to a missing license

inray.opcua.connection_manager.state

Describes the connection status of the OPC UA Connection Manager.

Value	Meaning	Practical Interpretation
0	Disconnected	No active connection to the OPC UA server
1	Connected	Connection established
2	InvalidConfig	Connection cannot be established due to an invalid configuration

---

Available Metrics

All metrics have the prefix inray. and can be enriched with attributes (dimensions) such as the flow name or the plug-in type.

Not all metrics are always available

. Plugin-specific metrics are exported only if the respective plug-in is licensed and active. Which metrics are actually provided depends on the configuration and runtime state of the OPC router.

Service & System

General status metrics for the OPC Router service and the host system.

Useful for:

• Monitoring the overall health of the service
• Detecting CPU, memory, or disk space bottlenecks
• Issuing alerts for critical operating conditions

These include, among others, the following metrics:

Metric	Type	Unit	Description
inray.service.escalation_level	Gauge	-	Global health status of the OPC router; the numerical values should be interpreted using the value table above
inray.service.used_memory_bytes	Gauge	Bytes	Memory usage of the process
inray.system.processor_usage	Gauge	%	CPU utilization of the host system
inray.system.memory_usage	Gauge	%	Host system memory usage
inray.system.free_memory_bytes	Gauge	Bytes	Available memory
inray.drives.{Drive}.free_bytes	Gauge	Bytes	Free disk space per drive
inray.drives.{Drive}.total_bytes	Gauge	Bytes	Total size per drive

Transfer Manager

Metrics for the OPC Router’s internal processing scheduler. These values form the basis of the Service Dashboard.

Useful for:

• Identifying bottlenecks in the processing path
• Assessing whether enough worker threads are available
• Analyzing whether increasing latencies are caused by load or by individual slow transfers

Recalculate the “System Load” display

The “System load” display in the service status is based directly on these two metrics:

• Regular: inray.transfer_manager.default_prio_average_workload
• High Prio: inray.transfer_manager.high_prio_average_workload

The calculation is performed in three steps:

1. For each worker thread, a workload sample is generated per measurement interval:

load_sample = activeMs / (activeMs + idleMs)

Here, activeMs is the time during which the thread is actually processing transfers, and idleMs is the time during which the thread is waiting.

2. The OPC Router stores the average of the last up to 10 samples for each worker thread:

load_thread = (load_sample_1 + load_sample_2 + ... + load_sample_n) / n

where n <= 10.

3. For the selected time period, the dashboard reads all stored values for the respective metric and calculates the arithmetic mean from them:

load_dashboard = (load_thread_1 + load_thread_2 + ... + load_thread_m) / m

The gauge is then displayed as a percentage:

Display in % = load_dashboard * 100

Example:

• A dashboard value of 0.37 corresponds to a displayed system load of 37 %.
• A dashboard value of 0.85 corresponds to 85% and indicates a high, sustained load.
• If no values are available for the selected time period, the gauge displays N/A.

Important for interpretation:

• The gauge does not show the host’s CPU utilization, but rather the utilization of the transfer workers over time.
• The gauge displays an averaged trend, not the current value of a single thread.
• For Regular, the values for normal and low priority are used; for High Prio, the values for high-priority workers are used.

These include, among others, the following metrics:

Metric	Type	Unit	Description
inray.transfer_manager.queue_depth	Gauge	Items	Number of transfer executions currently waiting across all priorities; a persistently rising value indicates a backlog
inray.transfer_manager.queue_depth.high_priority	Gauge	Items	Currently waiting high-priority executions; should ideally remain low
inray.transfer_manager.queue_depth.normal_priority	Gauge	Items	Currently pending normal-priority transfers
inray.transfer_manager.queue_depth.low_priority	Gauge	Items	Currently pending low-priority transfers
inray.transfer_manager.busy_worker_threads	Gauge	Threads	Busy worker threads
inray.transfer_manager.ready_worker_threads	Gauge	Threads	Ready worker threads
inray.transfer_manager.worker_thread_utilization	Gauge	%	Total utilization of all workers; Consistently high values combined with a growing queue indicate capacity limits
inray.transfer_manager.utilization.high_priority	Gauge	%	Percentage of processing time for high-priority transfers
inray.transfer_manager.utilization.normal_priority	Gauge	%	Percentage of processing time for normal-priority transfers
inray.transfer_manager.utilization.low_priority	Gauge	%	Percentage of processing time spent on low-priority transfers
inray.transfer_manager.average_triggers_per_second	Gauge	Items/s	Average input rate to the Transfer Manager
inray.transfer_manager.average_processing_time_per_item	Gauge	ms	Average processing time per processed item
inray.transfer_manager.trigger_transfer_delay	Histogram	ms	Time between trigger event and actual processing; important for latency analysis
inray.transfer_manager.tags_executed	Histogram	Tags	Number of tags processed per transfer; helps determine why a transfer is taking longer
inray.transfer_manager.default_prio_average_workload	Gauge	-	Backpressure for normal and low priority; the higher the value, the more work could not be taken on immediately
inray.transfer_manager.high_prio_average_workload	Gauge	-	Backpressure for high priority

Typical interpretation:

• queue_depth increases, worker_thread_utilization remains high: The router is continuously processing at its limit.
• average_processing_time_per_item increases, and tags_executed also increases: The transfers are becoming more complex, not just more numerous.
• high_prio_average_workload increases even though the normal queue is low: High-priority flows require targeted analysis.

Flows

Metrics per configured flow, dimensioned by the flow name.

Useful for:

• Identifying faulty or slow individual flows
• Comparing individual flows in Grafana or InfluxDB
• Triggering alerts for status changes or rising error rates

These include, among others, the following metrics:

Metric	Type	Unit	Description
inray.connections.execution.count	Counter	-	Total number of executions
inray.connections.execution.success.count	Counter	-	Successful executions
inray.connections.execution.error.count	Counter	-	Failed executions
inray.connections.execution.duration	Histogram	ms	Execution duration
inray.connections.execution.type.duration	Histogram	µs	Execution duration by result type, for example Ok, Error, or Aborted
inray.connections.state	Gauge	-	Current flow status; the numerical values should be interpreted using the value table above
inray.connections.execution.state	Gauge	-	Result of the last execution; useful for single-stat panels or status lights
inray.connections.total.ok.count	Gauge	-	Flows in the OK state
inray.connections.total.error.count	Gauge	-	Flows in an error state
inray.connections.total.deactivated.count	Gauge	-	Deactivated flows

Typical interpretation:

• execution.error.count increases while execution.count also increases: The flow is running but generating errors.
• execution.duration increases without an increased trigger load: The cause lies more in the processing or the target system than in the trigger.
• connections.state = 7: The flow is not experiencing a technical issue but is blocked due to licensing restrictions.

Triggers

Metrics for asynchronous triggers, dimensioned by flow name.

Useful for:

• Analyzing overflowing or undersized trigger queues
• Assessing whether triggers are being generated faster than they are processed
• Distinguishing between configuration errors and runtime bottlenecks

These include, among others, the following metrics:

Metric	Type	Unit	Description
inray.connections.trigger.async.status	Gauge	-	Status of the asynchronous trigger; the numerical values should be interpreted using the value table above
inray.connections.trigger.async.queued_events	Gauge	-	Currently queued trigger events that have not yet been processed
inray.connections.trigger.async.queue_limit	Gauge	-	Configured maximum size of the trigger queue
inray.connections.trigger.async.enqueue.count	Counter	-	Total number of events enqueued
inray.connections.trigger.async.dequeue.count	Counter	-	Total processed events
inray.connections.trigger.async.dequeue.dropped_timeout	Counter	-	Dropped events (timeout)
inray.connections.trigger.async.queued_per_second_avg	Gauge	Events/s	Average Inbound Rate
inray.connections.trigger.async.dequeued_per_second_avg	Gauge	Events/s	Average processing rate
inray.connections.trigger.async.dropped_per_second_avg	Gauge	Events/s	Average drop rate

Typical interpretation:

• queued_events is approaching queue_limit: The queue is too small or processing is too slow.
• queued_per_second_avg is consistently higher than dequeued_per_second_avg: A backlog is building up.
• dequeue.dropped_timeout is increasing: The flow cannot process trigger events in a timely manner.

OPC UA

OPC UA connection management metrics, dimensioned by the server address.

Useful for:

• Monitoring the stability of OPC UA connections
• Identifying configuration issues or unstable sessions
• Evaluating read and subscription behavior

These include, among others, the following metrics:

Metric	Type	Unit	Description
inray.opcua.connection_manager.state	Gauge	-	Connection status of the OPC UA manager; the numerical values should be interpreted using the value table above
inray.opcua.connection_manager.connection_established_successful.count	Counter	-	Successful connection establishment attempts
inray.opcua.connection_manager.connection_established_failed.count	Counter	-	Failed connection attempts
inray.opcua.connection_manager.keep_alive_failed.count	Counter	-	Keep-alive errors
inray.opcua.connection_manager.watchdog_cycle_duration	Histogram	ms	Duration of the watchdog cycle
inray.opcua.group.item_count	Gauge	-	Number of subscribed items per group

Plug-ins

Metrics for the plug-in system as well as plug-in-specific metrics for individual plug-in types. These metrics are only available if the respective plug-in is active and licensed.

General Plug-in Metrics

Metric	Type	Unit	Description
inray.plugins.instance.count	Gauge	-	Number of active plugin instances
inray.plugins.instance.status	Gauge	-	Status of a plugin instance; the numerical values should be interpreted using the value table above
inray.plugins.execution.count	Counter	-	Total number of plugin executions
inray.plugins.execution.time.histogram	Histogram	ms	Execution time per plugin instance
inray.plugins.total.ok.count	Gauge	-	Plugins in the OK state
inray.plugins.total.error.count	Gauge	-	Plug-ins in an error state
inray.plugins.total.warning.count	Gauge	-	Plug-ins with warnings

MQTT Plug-in

Metric	Type	Unit	Description
inray.mqtt.connections.active	Gauge	-	Active MQTT connections
inray.mqtt.connect.attempt	Counter	-	Connection attempts
inray.mqtt.connect.ok	Counter	-	Successful connections
inray.mqtt.connect.fail	Counter	-	Failed connections
inray.mqtt.publish.ok	Counter	-	Successfully published messages
inray.mqtt.publish.fail	Counter	-	Failed publications
inray.mqtt.publish.latency.ms	Histogram	ms	Publish latency
inray.mqtt.payload.bytes	Histogram	Bytes	Message size

.NET Runtime & Process

If system instrumentation is enabled, the OPC Router also exports standard metrics from the .NET runtime and the operating system process (enabled by default). These include, among other things, garbage collector statistics, thread pool utilization, and process CPU time.