Version: v2026.0.0

Monitoring and observability

ARender exposes metrics, health endpoints, and structured logs across all rendition services.

Actuator endpoints

All rendition services expose Spring Boot Actuator endpoints. The Document Service Broker exposes three endpoints by default:

management.endpoints.web.exposure.include=prometheus,metrics,health

The Document Converter, Document Renderer, and Document Text Handler expose an additional shutdown endpoint used by the broker's health check loop to restart failed services in standalone mode:

management.endpoints.web.exposure.include=prometheus,metrics,health,shutdown

These defaults and all metrics export settings (Prometheus, Elasticsearch, Datadog, CloudWatch) are documented in the Rendition properties — Shared metrics settings.

Available endpoints

Endpoint	Path	Purpose
Health	`/actuator/health`	Returns UP/DOWN status for the service
Metrics	`/actuator/metrics`	Lists all available metric names
Metrics (specific)	`/actuator/metrics/{name}`	Returns values for a specific metric
Prometheus	`/actuator/prometheus`	Prometheus scrape endpoint (disabled by default, see below)
Shutdown	`/actuator/shutdown`	Triggers graceful shutdown (converter, renderer, text handler only)

Enabling Prometheus

The Prometheus endpoint is disabled by default. To enable it, set:

management.endpoint.prometheus.access=unrestricted

Once enabled, Prometheus can scrape each service at /actuator/prometheus.

Metrics configuration

ARender uses Micrometer for metrics collection. Each service tags its metrics with a host identifier and application name.

Per-service host tags

Service	Host tag
Document Service Broker	`arender-broker`
Document Converter	`arender-taskconversion`
Document Renderer	`arender-jni`
Document Text Handler	`arender-pdfbox`

All services share the same application tag:

management.metrics.tags.application=arender

HTTP request distribution

Every service configures percentile histograms and SLA buckets for HTTP request metrics:

management.metrics.distribution.percentiles-histogram.http.server.requests=true
management.metrics.distribution.sla.http.server.requests=100ms, 400ms, 500ms, 2000ms
management.metrics.distribution.percentiles.http.server.requests=0.5, 0.9, 0.95, 0.99

Disabled default meters

By default, ARender disables several standard Micrometer meters to reduce noise. These are disabled across all services:

management.metrics.enable.tomcat=false
management.metrics.enable.http=false
management.metrics.enable.logback=false
management.metrics.enable.jvm=false
management.metrics.enable.process=false
management.metrics.enable.system=false
management.metrics.enable.application=false
management.metrics.enable.executor=false
management.metrics.enable.disk=false

To re-enable any of these, set the corresponding property to true. For example, to monitor JVM memory:

management.metrics.enable.jvm=true

ARender endpoint metrics

The broker and all services support fine-grained per-endpoint metrics. Each endpoint metric is disabled by default and can be enabled individually:

# Enable metrics for document loading
arender.endpoint.metrics.export.load.document.enabled=true

# Enable metrics for image rendering
arender.endpoint.metrics.export.image.enabled=true

# Enable metrics for text search
arender.endpoint.metrics.export.search.enabled=true

# Enable metrics for document conversion
arender.endpoint.metrics.export.convert.enabled=true

# Enable metrics for document comparison
arender.endpoint.metrics.export.compare.enabled=true

The full list of toggleable endpoints:

Property	Operation
`arender.endpoint.metrics.export.has.document.enabled`	Document existence check
`arender.endpoint.metrics.export.bookmarks.enabled`	Bookmark extraction
`arender.endpoint.metrics.export.document.layout.enabled`	Document layout retrieval
`arender.endpoint.metrics.export.load.document.content.enabled`	Document content loading
`arender.endpoint.metrics.export.get.file.chunk.enabled`	File chunk retrieval
`arender.endpoint.metrics.export.text.position.enabled`	Text position extraction
`arender.endpoint.metrics.export.document.annotation.enabled`	Annotation retrieval
`arender.endpoint.metrics.export.transformation.enabled`	Transformation orders
`arender.endpoint.metrics.export.document.metadata.enabled`	Metadata retrieval
`arender.endpoint.metrics.export.image.enabled`	Page image rendering
`arender.endpoint.metrics.export.page.contents.enabled`	Page content retrieval
`arender.endpoint.metrics.export.search.enabled`	Text search
`arender.endpoint.metrics.export.advanced.search.enabled`	Advanced text search
`arender.endpoint.metrics.export.load.document.enabled`	Document loading
`arender.endpoint.metrics.export.evict.enabled`	Cache eviction
`arender.endpoint.metrics.export.annotation.enabled`	Annotation operations
`arender.endpoint.metrics.export.compare.enabled`	Document comparison
`arender.endpoint.metrics.export.named.destination`	Named destination extraction
`arender.endpoint.metrics.export.weather.enabled`	System weather/status
`arender.endpoint.metrics.export.readiness.enabled`	Readiness check
`arender.endpoint.metrics.export.signature.enabled`	Signature verification
`arender.endpoint.metrics.export.printable.pdf.enabled`	Printable PDF generation
`arender.endpoint.metrics.export.convert.enabled`	Format conversion
`arender.endpoint.metrics.export.health.record.enabled`	Health record retrieval
`arender.endpoint.metrics.export.document.xfa.check.enabled`	XFA form detection

Metric tags

Control which tags are included in exported metrics:

# Tags to include (comma-separated)
arender.endpoint.metrics.export.whitelist.tags=host,mimeType

# Include correlation ID as a tag (adds cardinality, use with caution)
arender.endpoint.metrics.export.correlation.id.tag.enabled=false

# Tags to exclude from system meters
arender.system.metrics.export.blacklist.tags=

Metric collection mode

Choose between timing requests (with duration percentiles) or counting them:

# TIMER records duration; COUNTER records invocation count only
arender.metric.meter.tool=COUNTER

External monitoring systems

ARender supports exporting metrics to four external systems. Each is disabled by default.

Prometheus

Enable the Prometheus scrape endpoint as described above, then configure your Prometheus server to scrape each service:

prometheus.yml
scrape_configs:
  - job_name: 'arender-broker'
    metrics_path: '/actuator/prometheus'
    static_configs:
      - targets: ['broker-host:8761']
  - job_name: 'arender-converter'
    metrics_path: '/actuator/prometheus'
    static_configs:
      - targets: ['converter-host:19999']
  - job_name: 'arender-renderer'
    metrics_path: '/actuator/prometheus'
    static_configs:
      - targets: ['renderer-host:9091']
  - job_name: 'arender-text-handler'
    metrics_path: '/actuator/prometheus'
    static_configs:
      - targets: ['text-handler-host:8899']

Elasticsearch

management.elastic.metrics.export.enabled=true
management.elastic.metrics.export.step=5m
management.elastic.metrics.export.index=arender-micrometer-metrics
management.elastic.metrics.export.host=http://localhost:9200

Datadog

management.datadog.metrics.export.enabled=true
management.datadog.metrics.export.api-key=YOUR_KEY
management.datadog.metrics.export.step=5m
management.datadog.metrics.export.uri=https://app.datadoghq.com/

CloudWatch

management.cloudwatch.metrics.export.enabled=true
management.cloudwatch.metrics.export.namespace=brokerNameSpace
management.cloudwatch.metrics.export.step=5m
management.cloudwatch.metrics.export.batchSize=20
management.cloudwatch.metrics.export.region=eu-west-1

Each service uses its own namespace value. The defaults are brokerNameSpace, converterNameSpace, rendererNameSpace, and pdfboxNameSpace.

Health checks and probes

Kubernetes liveness and readiness probes

The Helm chart configures HTTP probes for all services:

Service	Liveness path	Readiness path	Liveness delay	Readiness delay	Period
Document Service Broker	`/swagger-ui/index.html`	`/health/readiness`	30s	60s	15s
Document Converter	`/actuator/health`	`/health/readiness`	30s	60s	15s
Document Renderer	`/actuator/health`	`/health/readiness`	30s	60s	15s
Document Text Handler	`/actuator/health`	`/health/readiness`	30s	60s	15s

These values are configurable per service in values.yaml:

values.yaml
rendition:
  broker:
    deployment:
      livenessProbe:
        initialDelaySeconds: 30
        periodSeconds: 15
        timeoutSeconds: 3
      readinessProbe:
        initialDelaySeconds: 60
        periodSeconds: 15
        timeoutSeconds: 1

Custom liveness and readiness paths can be set:

values.yaml
rendition:
  broker:
    deployment:
      livenessProbe:
        path: "/actuator/health"
      readinessProbe:
        path: "/health/readiness"

Internal health monitoring

The broker runs an internal health check for each registered microservice. This is separate from Kubernetes probes. The broker pings each service's health endpoint and tracks their status internally:

If a service is detected as DOWN and health.check.restart.enabled=true (standalone mode), the broker sends a POST to /actuator/shutdown and restarts the process.

Log configuration

ARender uses Logback for logging. There are two logging configurations: one for Docker images (jib-packaged) and one for local Spring Boot development.

Docker/Kubernetes log pattern

In the Helm chart, the logging ConfigMap generates a Logback configuration that routes log levels to stdout/stderr and optionally to rolling files:

values.yaml
rendition:
  logging:
    default:
      consoleOnly: false
      logLevels:
        business: info      # com.arondor.arender classes
        technical: warn     # Spring, Tomcat, Jetty classes
      display:
        date: true
        podName: true

Set consoleOnly: true to disable file appenders entirely and output everything to the console, which is the recommended approach for container environments where a log aggregator collects stdout/stderr.

Log files

When file logging is enabled, each service writes to /arender/logs/:

Service	Log file	Additional log files
Document Service Broker	`arender-broker.log`	`arender-perf.log`, `arender-health.log`
Document Converter	`arender-converter.log`
Document Renderer	`arender-renderer.log`
Document Text Handler	`arender-handler.log`

All log files use a FixedWindowRollingPolicy with a max size of 2 MB per file (Helm) or 50 MB (Docker image default), compressed with ZIP, up to 50 archived files.

Performance log

The broker writes a dedicated performance log via the LoggerInterceptor class. This log tracks per-request timing:

logger name: com.arondor.viewer.common.logger.LoggerInterceptor

Health log

The broker's health check activity is logged separately:

logger name: com.arondor.arender.micro.services.rendition.jobs.MicroServiceHealthCheckJob

Persistent log storage (Kubernetes)

To persist logs across pod restarts, enable log persistence in the Helm chart:

values.yaml
rendition:
  logging:
    persistance:
      enabled: true
      storage:
        size: 1Gi
      accessModes: "ReadWriteMany"

Changing log levels at runtime

To change a service's log level without restarting, mount a custom logback.xml via the Helm chart:

values.yaml
rendition:
  broker:
    logging:
      useDefault: false
      custom: |
        <?xml version="1.0" encoding="UTF-8"?>
        <configuration>
          <appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
            <encoder>
              <pattern>%date %level [%thread] %logger{10} [%file:%line] %msg%n</pattern>
            </encoder>
          </appender>
          <logger name="com.arondor.arender" level="DEBUG" />
          <root level="info">
            <appender-ref ref="STDOUT" />
          </root>
        </configuration>

Key metrics to monitor

These metrics are the most useful for tracking ARender health and performance in production.

System-level

What to monitor	How
Service availability	Kubernetes liveness/readiness probes, or broker health check status
Pod restarts	`kube_pod_container_status_restarts_total` (from kube-state-metrics)
CPU and memory usage	Standard container metrics from cAdvisor or node-exporter
Shared volume usage	Disk usage on the `/arender/tmp` PVC

Application-level

What to monitor	Metric / indicator
HTTP request latency (p50, p95, p99)	`http.server.requests` percentile histogram
HTTP request error rate	`http.server.requests` filtered by status 5xx
Conversion duration	Enable `arender.endpoint.metrics.export.convert.enabled=true`
Rendering duration	Enable `arender.endpoint.metrics.export.image.enabled=true`
Search duration	Enable `arender.endpoint.metrics.export.search.enabled=true`
Document load count	Enable `arender.endpoint.metrics.export.load.document.enabled=true`

Common operational issues

Microservice not discovered by broker

Symptoms: Viewer shows an error when loading a document. Broker logs: "Found 0 instance of document-converter".

Causes:

The microservice container is not running or has not started yet.
In Docker Compose: the DSB_KUBEPROVIDER_KUBE.HOSTS_* environment variables do not match the service hostname.
In Kubernetes: the service DNS name in the broker ConfigMap does not resolve. Check that the target service's Kubernetes Service object exists in the correct namespace.

Diagnosis:

# Check broker logs for discovery messages
kubectl logs deployment/arender-rendition-broker | grep "Found 0 instance"

# Verify DNS resolution from the broker pod
kubectl exec deployment/arender-rendition-broker -- nslookup arender-rendition-converter

Shared volume file-not-found errors

Symptoms: Conversion succeeds but rendering fails. Broker logs: "FileNotFoundException" referencing /arender/tmp/....

Causes:

The shared PVC is not mounted in all pods.
Different PVCs are used across services (check claimName consistency).
The storage class does not actually support ReadWriteMany.

Diagnosis:

# Verify all pods mount the same PVC
kubectl get pods -o jsonpath='{range .items[*]}{.metadata.name}{"\t"}{.spec.volumes[*].persistentVolumeClaim.claimName}{"\n"}{end}'

High conversion latency

Symptoms: Office or email documents take a long time to render the first page.

Causes:

LibreOffice is slow to start its first conversion. Subsequent conversions are faster.
The converter pod is CPU-constrained. Check resource limits.
Large documents or spreadsheets with many cells exceed practical limits. The excel.maximum.cell.count property (default 1,000,000) caps spreadsheet size.

Diagnosis:

Enable converter endpoint metrics: arender.endpoint.metrics.export.convert.enabled=true
Check converter logs for timeout messages.
Review pod CPU usage during conversion.

PDFOwl process pool exhaustion

Symptoms: Rendering requests time out. Renderer logs: watchdog timeout messages.

Causes:

All PDFOwl processes are busy with long-running renders.
A document with very high-resolution or many layers exhausts the memory limit.

Remediation:

Increase the memory limit: pdfowl.memlimit.mb=2048
Increase the watchdog timeout: pdfowl.client.watchdog=20000
Scale renderer replicas.

Actuator endpoints​

Available endpoints​

Enabling Prometheus​

Metrics configuration​

Per-service host tags​

HTTP request distribution​

Disabled default meters​

ARender endpoint metrics​

Metric tags​

Metric collection mode​

External monitoring systems​

Prometheus​

Elasticsearch​

Datadog​

CloudWatch​

Health checks and probes​

Kubernetes liveness and readiness probes​

Internal health monitoring​

Log configuration​

Docker/Kubernetes log pattern​

Log files​

Performance log​

Health log​

Persistent log storage (Kubernetes)​

Changing log levels at runtime​

Key metrics to monitor​

System-level​

Application-level​

Common operational issues​

Microservice not discovered by broker​

Shared volume file-not-found errors​

High conversion latency​

PDFOwl process pool exhaustion​

Related pages​

Actuator endpoints

Available endpoints

Enabling Prometheus

Metrics configuration

Per-service host tags

HTTP request distribution

Disabled default meters

ARender endpoint metrics

Metric tags

Metric collection mode

External monitoring systems

Prometheus

Elasticsearch

Datadog

CloudWatch

Health checks and probes

Kubernetes liveness and readiness probes

Internal health monitoring

Log configuration

Docker/Kubernetes log pattern

Log files

Performance log

Health log

Persistent log storage (Kubernetes)

Changing log levels at runtime

Key metrics to monitor

System-level

Application-level

Common operational issues

Microservice not discovered by broker

Shared volume file-not-found errors

High conversion latency

PDFOwl process pool exhaustion

Related pages