[receiver/kafka] otelcol_kafka_receiver_offset_lag not calculated correctly

[dyanneo]

Component(s)

receiver/kafka

What happened?

We wanted to collect, graph, and alert on lag for the kafka receiver, but observed unexpected behavior when observing the otelcol_kafka_receiver_offset_lag's last values, compared to the values observed using kafka's consumer-groups utility.

Description

The value of last for the measurement otelcol_kafka_receiver_offset_lag does not appear to be calculated correctly.
Also, for context, we are seeing an issue in the otel collector where it keeps emitting lag metrics for partitions it's no longer consuming.

• We understand that topic is not present as a tag in this metric, per issue 35336

Steps to Reproduce

• We are emitting and collecting the otelcol_kafka_receiver_offset_lag metrics in our collector deployments
• We are using grafana to view this metric
• Upon querying for this metric's last value, we were surprised to see values that didn't correspond to kafka utility outputs.

Expected Result

Per this screen shot, the value of partition 4's lag as shown using kafka's consumer-groups.sh utility changes over time, and is in the low hundreds or smaller:

• you can see starting and ending offsets and the calculated lag of 117, 74, 112, and 129

Actual Result

Per this screen shot, the value of partition 4's lag does not match what true lag is, per kafka's tools:

Query in Grafana query builder:

Query as Grafana is running it:

Collector version

otelcol_version: 0.109.0

Environment information

Environment

OS: linux
container

OpenTelemetry Collector configuration

---
receivers:
  otlp:
    protocols:
      http:
        endpoint: 0.0.0.0:PORT
  prometheus:
    config:
      scrape_configs:
        - job_name: otel-collector-metrics
          scrape_interval: 30s
          static_configs:
            - targets: ["127.0.0.1:PORT"]

processors:
  memory_limiter:
    check_interval: 1ms
    limit_percentage: 70
    spike_limit_percentage: 15

  resource:
    attributes:
      - key: application
        from_attribute: service.name
        action: insert

  resource/internal:
    attributes:
      - key: labels.application
        value: ${env:CLOUD_APPLICATION}
        action: upsert
      - key: application
        value: ${env:CLOUD_APPLICATION}
        action: upsert
      
  filter/application:
    error_mode: ignore
    traces:
      span:
        - 'resource.attributes["application"] == nil'
    metrics:
      metric:
        - 'resource.attributes["application"] == nil'

  filter/tracingsampler:
    error_mode: ignore
    traces:
      span:
        - 'attributes["sampler.parent"] == nil'

  filter/summarytype:
    error_mode: ignore
    metrics:
      metric:
        - "type == METRIC_DATA_TYPE_SUMMARY"

  metricstransform/version_correction:
    transforms:
      - match_type: strict
        include: tgt.otel.extension.heartbeat
        action: update
        operations:
          - action: update_label
            label: otel.extension.version
            value_actions:
              - value: "v0.8.0\n"
                new_value: "v0.8.0"
          - action: update_label
            label: otel.scope.version
            value_actions:
              - value: "v0.8.0\n"
                new_value: "v0.8.0"

  batch/metrics:
    timeout: 5s
    send_batch_size: 100
    send_batch_max_size: 101

  batch/traces:
    timeout: 5s
    send_batch_size: 33
    send_batch_max_size: 34

exporters:
  tgtkafkaexporter:
    brokers:
      - BROKER_REDACTED
    topic: TOPIC_REDACTED
    protocol_version: 2.7.0
    encoding_extension: tgtinflux_encoding
    producer:
      compression: snappy
      max_message_bytes: 2000000

  kafka:
    brokers:
      - BROKER_REDACTED
    topic: TOPIC_REDACTED
    protocol_version: 2.7.0
    producer:
      compression: snappy
      max_message_bytes: 2000000

service:
  pipelines:
    metrics:
      receivers:
        - otlp
      processors:
        - memory_limiter
        - resource
        - filter/application
        - filter/summarytype
        - metricstransform/version_correction
        - batch/metrics
      exporters:
        - tgtkafkaexporter

    metrics/internal:
      receivers:
        - prometheus
      processors:
        - resource
        - resource/internal
        - filter/application
        - filter/summarytype
        - metricstransform/version_correction
        - batch/metrics
      exporters:
        - tgtkafkaexporter

    traces:
      receivers:
        - otlp
      processors:
        - memory_limiter
        - resource
        - filter/application
        - filter/tracingsampler
        - batch/traces
      exporters:
        - kafka

  telemetry:
    logs:
      level: "info"
    metrics:
      address: 0.0.0.0:PORT

Log output

***WIP***

Additional context

We think the reason for the incorrect data is the fact that the gauge exists within OTEL's registry even after a rebalance and the metric is not receiving updates

Where this gauge is defined:

opentelemetry-collector-contrib/receiver/kafkareceiver/internal/metadata/generated_telemetry.go

Lines 75 to 79 in 0d28558

	builder.KafkaReceiverOffsetLag, err = builder.meters[configtelemetry.LevelBasic].Int64Gauge(
	"otelcol_kafka_receiver_offset_lag",
	metric.WithDescription("Current offset lag"),
	metric.WithUnit("1"),
	)

One of the places where this gauge is updated:

opentelemetry-collector-contrib/receiver/kafkareceiver/kafka_receiver.go

Line 560 in 0d28558

c.telemetryBuilder.KafkaReceiverOffsetLag.Record(ctx, claim.HighWaterMarkOffset()-message.Offset-1, metric.WithAttributeSet(attrs))

Where this could be addressed:

opentelemetry-collector-contrib/receiver/kafkareceiver/kafka_receiver.go

Lines 529 to 531 in 0d28558

	func (c *tracesConsumerGroupHandler) Cleanup(session sarama.ConsumerGroupSession) error {
	c.telemetryBuilder.KafkaReceiverPartitionClose.Add(session.Context(), 1, metric.WithAttributes(attribute.String(attrInstanceName, c.id.Name())))
	return nil

[github-actions]

Pinging code owners:

• receiver/kafka: @pavolloffay @MovieStoreGuy

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[StephanSalas]

Hi @dyanneo,

Based on my initial analysis:

Cleanup is run at the end of a session, once all ConsumeClaim goroutines have exited but before the offsets are committed for the very last time.
Cleanup(ConsumerGroupSession) error

We have to be potentially careful here with using Cleanup func to solve this issue. By the time Cleanup is called, the session's claims may no longer reflect the partitions that were assigned, or session.Claims() might be empty. We may need to check that assumption as well...

That said... I agree with your approach as long as we are guaranteed on Cleanup(), that the session is accurately populated, we can do something like this:

for topic, partitions := range session.claims() {
    for _, partition := range partitions {
        c.telemetryBuilder.KafkaReceiverPartitionClose.Add(session.Context(), 1, metric.WithAttributes(
            attribute.String(attrInstanceName, c.id.Name()),
            attribute.String(attrTopic, topic),
            attribute.String(attrPartition, strconv.Itoa(int(partition))),
        ))
        // add cleanup for _offset_lag_ metric here as well.
    }
}

[dyanneo]

@StephanSalas Thanks for your inputs on this issue. I appreciate your concerns and the suggestion makes sense to me.