apache/druid · critical · IllegalStateException

Unknown metric value type

Error message

Unknown metric value type: %s

What it means

IncrementalIndex throws this ISE when building a metric column selector for a metric whose ValueType is not LONG, FLOAT, or DOUBLE. It is an internal guard in factRowSelector/makeMetricColumnSelector: only numeric metric types are supported in the ingestion index. Encountering it means a metric column resolved to an unexpected type.

Solutions

1. Fix the metricsSpec so every metric is LONG, FLOAT, or DOUBLE (e.g. change 'string' to 'long' or wrap with the correct aggregator)
2. Check the input field actually contains numeric values and remove it from metricsSpec if it is a dimension
3. Upgrade Druid if a new ValueType was added upstream and this build predates its selector

Example fix

// before
"metricsSpec": { "metrics": [ { "type": "stringFirst", "name": "tags", "fieldName": "tags" } ] }
// after
"metricsSpec": { "metrics": [ { "type": "longSum", "name": "count", "fieldName": "count" } ] }

for (AggregatorFactory f : spec.getMetricsSpec()) {
  String t = f.getTypeName();
  if (!(t.equals("long") || t.equals("float") || t.equals("double"))) {
    throw new IllegalArgumentException("Unsupported metric type: " + t);
  }
}

try { index.add(row); } catch (ISE e) { if (e.getMessage().startsWith("Unknown metric value type")) { /* halt and fix metricsSpec */ } throw e; }

When it happens

Trigger: Ingesting rows into an IncrementalIndex (realtime/Kafka ingestion) where a metric configured in the metricsSpec has an unsupported ValueType (e.g. STRING or an unknown type) so the switch in factRowSelector falls through to default.

Common situations: Metrics spec mistakes: declaring a dimension as a metric, using a string-typed field as a metric, or schema/API version changes that introduce new ValueTypes not handled by the running Druid version.

Related errors

• incrementIndexSchema cannot be null
• String is not a metric column type
• Aggregation [ ] does not support column [ ] of type [ ]…
• An inline input source must provide one or more columns
• Attempt to add row to swapped-out sink for segment

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/cc61152a70554cf8. Report an issue: GitHub.

Appendix: source

Thrown at processing/src/main/java/org/apache/druid/segment/incremental/IncrementalIndex.java:795

  {
    final MetricDesc metricDesc = rowSelector.getMetric(metric);
    if (metricDesc == null) {
      return NilColumnValueSelector.instance();
    }
    int metricIndex = metricDesc.getIndex();
    switch (metricDesc.getCapabilities().getType()) {
      case COMPLEX:
        return new ObjectMetricColumnSelector(rowSelector, currEntry, metricDesc);
      case LONG:
        return new LongMetricColumnSelector(rowSelector, currEntry, metricIndex);
      case FLOAT:
        return new FloatMetricColumnSelector(rowSelector, currEntry, metricIndex);
      case DOUBLE:
        return new DoubleMetricColumnSelector(rowSelector, currEntry, metricIndex);
      case STRING:
        throw new IllegalStateException("String is not a metric column type");
      default:
        throw new ISE("Unknown metric value type: %s", metricDesc.getCapabilities().getType());
    }
  }

  public Interval getInterval()
  {
    DateTime min = DateTimes.utc(minTimestamp);
    return new Interval(min, isEmpty() ? min : queryGranularity.increment(DateTimes.utc(getMaxTimeMillis())));
  }

  @Nullable
  public DateTime getMinTime()
  {
    return isEmpty() ? null : DateTimes.utc(getMinTimeMillis());
  }

  @Nullable
  public DateTime getMaxTime()
  {

View on GitHub (pinned to 9b90983fd2)