Monitoring

Jadice flow provides application metrics which can be observed.

Configuration

To enable metrics, use the appropriate flag in the helm charts or the application.yaml:

# Metrics
management:
  metrics:
    enable:
      all: true
    export.prometheus.enabled: true
    export.influx.enabled: false
    export.jmx.enabled: false
  endpoints:
    web.exposure.include: health,info,bindings,metrics,prometheus

This example enables the prometheus exposure of the metrics, but not InfluxDB or JMX.

Access metrics

Prometheus

If prometheus is enabled, the metrics will be propagated under the http://localhost:8080/actuator/prometheus endpoint and is scraped by the Prometheus server.

The metrics can be visualized, e.g. with Grafana by adding the prometheus adapter.

JMX

The JMX exposure exposes the metrics via a default Java MBean server handling the jmx request. Depending on the JVM Vendor there are specific Java Mission Control Derivates which can be used to access the JVM and read the metrics. By default, most JDKs come with the jconsole-Tool, which provides the most basic access to the metrics.

InfluxDB

When using influx, application metrics are sent from the controller to the InfluxDB.

Configuration:

# Metrics
management:
  metrics:
    enable.all: true
    export.prometheus.enabled: false
    export.influx.enabled: true
    export.jmx.enabled: true
    export:
      influx:
        api-version: v2
        bucket: ${jf.influxdb.bucket} # Specifies the destination bucket for writes
        org: my-org #Specifies the destination organization for writes.
        token: ${jf.influxdb.token}
        uri: ${jf.influxdb.uri} # The URI for the Influx backend. (Default: http://localhost:8086/api/v2)
        # compressed: true # Whether to enable GZIP compression of metrics batches published to Influx. (Default: true)
        # enabled: true # Whether exporting of metrics to this backend is enabled. (Default: true)
        step: 5s # Step size (i.e. reporting frequency) to use. (Default: 1m)
        # connect-timeout: 1s # Connection timeout for requests to this backend. (Default: 1s)
        # read-timeout: 10s # Read timeout for requests to this backend. (Default: 10s)
  endpoints:
    web.exposure.include: health,info,bindings,metrics,prometheus

The metrics can then be viewed e.g. directly in the InfluxDB Web App or they can be visualized with Grafana like before, by adding an InfluxDB Adapter.

Metrics overview

The jadice flow controller will provide basic application metrics like CPU usage, JVM memory and the like via the default actuator.

In addition, several jadice flow application metrics are available to get more information:

Metric name	Type	Tags	Description
jadice-flow.step.timer	Timer	fullStepName (canonical step name)	Time for a step.
jadice-flow.itemProcessor.part.added	Counter	fullStepName (canonical step name)	Added part count for a step
jadice-flow.jobLauncher.runningJobs	Gauge		Total amount of running jobs
jadice-flow.jobLauncher.queuedJobs	Gauge		Total amount of queued jobs
jadice-flow.jobLauncher.queuedJobsByType	Gauge	job (Job name)	The amount of queued jobs per JobTemplate

Tracing

The controller writes application traces which can be accessed later on using jaegertracing.

Configuration

To enable tracing, use the appropriate flag in the helm charts or the application.yaml:

# Tracing
opentracing:
  jaeger:
    enabled: true
    udp-sender.host: localhost
    udp-sender.port: 6831
    log-spans: false
    probabilistic-sampler.sampling-rate: 1

A Trace-ID will be provided in each job. E.g. when inspecting the Job in the UI, the trace-ID can be copied and entered in the Tracing UI to show the job trace.

Access traces

The traces can be accessed in the Tracing UI. The jaeger UI is linked in the jadice flow UI, linking to the jaeger UI URL (e.g. http://localhost:16686/). The URL is configured in the UI's application.yaml.

Available Worker Metrics

Analyzer Worker metrics

metric	description	labels (examples)
jadice_server_analyzer_mimetype	number of incoming documents per mime type	type="image/tiff"
jadice_server_analyzer_mimetype_before	number of incoming documents per mime type (before analysis)	type="image/tiff"
jadice_server_analyzer_mimetype_after	number of processed documents per mime type (after analysis)	type="text/plain"

Decompress Worker

metric	description	labels (examples)
jadice_server_decompress_mimetype	number of processed documents per mime type	type="application/x-rar-compressed"

Libre Office Worker

metric	description	labels (examples)
jadice_server_libreoffice_mimetype	number of incoming documents per mime type	type="application/msword"
jadice_server_libreoffice_mimetype_total	number of incoming documents per mime type	type="application/msword"
jadice_invocation	number of incoming Jobs for this worker

OCR Worker

metric	description	labels (examples)
jadice_server_ocr_mimetype	number of incoming documents per mime type	type="image/tiff"
jadice_server_ocr_mimetype_total	number of processed documents per mime type	type="image/tiff"
jadice_server_ocr_pagesegmentationmode	number of processed documents per page segmentation mode	mode="3"

Grafana

Grafana is one of many supported monitoring solutions.

Pre-configured Grafana dashboards are supposed to be delivered with the Helm Charts as part of the products.

Health

A standard actuator health check is provided via /actuator/health/liveness.

Sample response for GET /actuator/health/liveness

{
    "status": "UP"
}