Ohh no! UnusedConnectionTimeout

Error message

Ohh no! UnusedConnectionTimeout[%s] is longer than readTimeout[%s], please correct the configuration, this might not be supported in future.

What it means

Log warning in DruidHttpClientConfig.getUnusedConnectionTimeout. If unusedConnectionTimeout is greater than or equal to readTimeout, the pool's idle-connection eviction is ineffective because a connection would be closed by the read timeout before it can be evicted for idleness; the config warns that this combination may be unsupported in future versions.

Solutions

Set druid.http.unusedConnectionTimeout strictly less than druid.http.readTimeout (e.g. unusedConnectionTimeout=PT4M with readTimeout=PT5M).
Or unset unusedConnectionTimeout (null) so only readTimeout applies.
Double-check the durations are not accidentally swapped in runtime.properties.
Plan for the future version where this combination may be rejected outright.

Example fix

// before
druid.http.readTimeout=PT5M
druid.http.unusedConnectionTimeout=PT10M
// after
druid.http.readTimeout=PT5M
druid.http.unusedConnectionTimeout=PT4M

Defensive patterns

Strategy: validation

Validate before calling

// Validate http timeout ordering before building the client
Duration unused = config.getUnusedConnectionTimeoutConfig();
Duration read = config.getReadTimeoutConfig();
if (unused != null && read != null && unused.compareTo(read) >= 0) {
  throw new ConfigException("unusedConnectionTimeout must be < readTimeout");
}

Prevention

Keep unusedConnectionTimeout strictly below readTimeout
Validate duration ordering in config linters
Unset unusedConnectionTimeout if readTimeout alone suffices
Check for swapped values when copying configs between environments

When it happens

Trigger: getUnusedConnectionTimeout() is called (during HttpClient builder construction) with both druid.http.unusedConnectionTimeout and druid.http.readTimeout set and unusedConnectionTimeout >= readTimeout, e.g. unusedConnectionTimeout=PT10M with readTimeout=PT5M.

Common situations: Operators setting a long unusedConnectionTimeout to keep pooled connections alive while a readTimeout was added for query latency; copied http client tuning between node types; typo swapping the two duration values.

Understand the failure class

Background: Conflicting config options: "cannot be used together" — configuration validation errors across open-source libraries — this error's family across 162 libraries.

Timeouts: ETIMEDOUT, deadlines, and hung requests — what actually expires when a request times out.

Related errors

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/0c2e5d83f2e8c843. Report an issue: GitHub.

Appendix: source

Thrown at server/src/main/java/org/apache/druid/guice/http/DruidHttpClientConfig.java:113

  {
    return compressionCodec;
  }

  public int getNumRequestsQueued()
  {
    return numRequestsQueued;
  }

  public int getRequestBuffersize()
  {
    return requestBuffersize;
  }

  public Duration getUnusedConnectionTimeout()
  {
    if (unusedConnectionTimeout != null && readTimeout != null
        && unusedConnectionTimeout.toStandardDuration().compareTo(readTimeout.toStandardDuration()) >= 0) {
      LOG.warn(
          "Ohh no! UnusedConnectionTimeout[%s] is longer than readTimeout[%s], please correct"
          + " the configuration, this might not be supported in future.",
          unusedConnectionTimeout,
          readTimeout
      );
    }
    return unusedConnectionTimeout == null ? null : unusedConnectionTimeout.toStandardDuration();
  }

  public long getMaxQueuedBytes()
  {
    return maxQueuedBytes.getBytes();
  }

  public boolean isEagerInitialization(boolean defaultValue)
  {
    if (null == eagerInitialization) {
      return defaultValue;

View on GitHub (pinned to 9b90983fd2)