apache/hadoop · error · RuntimeException

CumulativeCpuUsageEmulatorPlugin got interrupted. Exiting.

Error message

CumulativeCpuUsageEmulatorPlugin got interrupted. Exiting.

What it means

In CumulativeCpuUsageEmulatorPlugin's midlock-adjustment loop, the plugin calls emulatorCore.compute() and Thread.sleep(100) until CPU usage reaches the progress-weighted target. If the thread is interrupted during that sleep it converts the InterruptedException into a RuntimeException with this message. Interruption here is Hadoop's normal mechanism for cancelling a task attempt, so this usually means the simulated task was killed or aborted, not that CPU emulation is broken.

Source

Thrown at hadoop-tools/hadoop-gridmix/src/main/java/org/apache/hadoop/mapred/gridmix/emulators/resourceusage/CumulativeCpuUsageEmulatorPlugin.java:285

        long projectedUsage = 
          currentCpuUsage + (long)((1 - currentProgress) * rate);
        
        if (projectedUsage < targetCpuUsage) {
          // determine the correction factor between the current usage and the
          // expected usage and add some weight to the target
          long currentWeighedTarget = 
            (long)(targetCpuUsage 
                   * getWeightForProgressInterval(currentProgress));
          
          while (getCurrentCPUUsage() < currentWeighedTarget) {
            emulatorCore.compute();
            // sleep for 100ms
            try {
              Thread.sleep(100);
            } catch (InterruptedException ie) {
              String message = 
                "CumulativeCpuUsageEmulatorPlugin got interrupted. Exiting.";
              throw new RuntimeException(message);
            }
          }
        }
        
        // set the last seen progress
        lastSeenProgress = progress.getProgress();
        // set the last seen usage
        lastSeenCpuUsage = getCurrentCPUUsage();
      }
    }
  }

  @Override
  public void initialize(Configuration conf, ResourceUsageMetrics metrics,
                         ResourceCalculatorPlugin monitor,
                         Progressive progress) {
    this.monitor = monitor;
    this.progress = progress;

View on GitHub (pinned to 2add963021)

Solutions

Treat the exception as expected cancellation when you intentionally killed the job; no fix needed for the simulation itself
Raise task timeouts (mapreduce.task.timeout) so CPU-emulating tasks are not interrupted by the framework mid-emulation
If you maintain this code, restore the interrupt flag and throw a quieter checked path instead of RuntimeException

Defensive patterns

Strategy: try-catch

Try / catch

try {
  plugin.middleware... // run job with CPU emulation
} catch (RuntimeException e) {
  if (e.getMessage() != null && e.getMessage().contains("got interrupted")) {
    // task was cancelled (kill/abort); treat as shutdown, restore interrupt flag
    Thread.currentThread().interrupt();
    return;
  }
  throw e;
}

Prevention

Raise mapreduce.task.timeout above the longest CPU-emulation task so framework interrupts do not fire mid-emulation
Expect these exceptions when killing gridmix runs — filter them in log analysis rather than debugging them

When it happens

Trigger: Killing/aborting the gridmix job or task attempt while a CPU-emulation map task is inside its 100ms sleep; task timeout leading the framework to interrupt the task thread; cluster shutdown interrupting running simulated tasks.

Common situations: 'hadoop job -kill' or YARN application kill during a gridmix run with resource emulation enabled (gridmix.emulators.resource-usage.plugins including the CPU plugin); long CPU-emulation tasks exceeding task timeouts; JVM reuse scenarios where interrupts leak between tasks.

Related errors

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/4275e0813f0b0a48. Report an issue: GitHub.