OpenTelemetry Collector Pipeline Health Dashboard Template

Use this dashboard to monitor your OpenTelemetry Collector instances across receiver throughput, processor drop rates, exporter delivery, queue depth, and process resource usage.

What This Dashboard Monitors

• Overview: Spans, metric points, and log records received and sent per second at a glance
• Receivers: Accepted and refused/failed signals per second, broken down by receiver
• Processors: Items entering and leaving each processor, batch send size percentiles, and timeout-triggered flushes
• Exporters: Sent and failed signals by exporter, queue size vs. capacity, queue utilization percentage, and in-flight requests
• Process Resources: Heap memory, RSS memory, CPU usage, and memory allocation rate per Collector instance

Metrics Included

Overview

• Spans Received /s: Spans entering the pipeline per second across all receivers
• Metric Points Received /s: Metric data points entering the pipeline per second
• Log Records Received /s: Log records entering the pipeline per second
• Spans Sent /s: Spans delivered to the backend per second across all exporters
• Metric Points Sent /s: Metric data points delivered to the backend per second
• Log Records Sent /s: Log records delivered to the backend per second

Receivers

• Accepted Spans /s by Receiver: Spans entering the pipeline per second, by receiver
• Refused & Failed Spans /s by Receiver: Refused spans from pipeline back-pressure and failed spans from receiver errors, grouped by receiver. Investigate any non-zero value.
• Accepted Metric Points /s by Receiver: Metric points entering the pipeline per second, by receiver
• Refused & Failed Metric Points /s by Receiver: Refused and failed metric points by receiver
• Accepted Log Records /s by Receiver: Log records entering the pipeline per second, by receiver
• Refused & Failed Log Records /s by Receiver: Refused and failed log records by receiver

Processors

• Items Incoming /s by Processor: Signals entering each processor per second
• Items Outgoing /s by Processor: Signals leaving each processor per second. A rate below incoming means the processor is dropping or filtering data.
• Batch Send Size (p50/p95/p99): Items per batch at three percentiles. A large gap between p50 and p99 points to bursty traffic.
• Batch Timeout Trigger Sends /s by Processor: Batches flushed by timeout per second by processor. A high rate with small batch sizes means the configured batch size is too large for current traffic.

Exporters

• Spans Sent /s by Exporter: Spans delivered to the backend per second, by exporter
• Span Send Failures /s by Exporter: Spans the exporter failed to deliver per second. Investigate any non-zero value.
• Metric Points Sent /s by Exporter: Metric points delivered per second, by exporter
• Metric Point Send Failures /s by Exporter: Metric points the exporter failed to deliver per second
• Log Records Sent /s by Exporter: Log records delivered per second, by exporter
• Log Record Send Failures /s by Exporter: Log records the exporter failed to deliver per second
• Exporter Queue Size vs Capacity: Queue depth vs. capacity per exporter. Size approaching capacity means the exporter cannot keep up with incoming data.
• Exporter Queue Utilization %: Queue fill percentage per exporter. Above 80% the exporter risks dropping data under sustained load.
• Exporter In-Flight Requests by Exporter: Active export requests including retries. High values alongside a slow-draining queue point to backend latency or connectivity issues.

Process Resources

• Heap Memory Allocated: Heap bytes held by live objects per Collector instance. Sustained growth between GC cycles points to a memory leak.
• Process RSS Memory: Physical memory per Collector instance, including Go runtime overhead
• CPU Usage (user + system): CPU seconds consumed per second per instance. Values near the available core count indicate CPU saturation.
• Memory Allocation Rate: Heap allocation throughput in bytes per second per instance. High rates increase GC pressure and CPU overhead.

Dashboard Variables

• service_name: Filter by Collector service name

Related Dashboards

• Host Metrics (VM)
• Kubernetes Node Metrics
• FluxCD
• KEDA