nodejs/node · error · Exception

Could not set CPU governor. Present value is %s

Error message

Could not set CPU governor. Present value is %s

What it means

Thrown by CustomMachineConfiguration.SetCPUGovernor in run_perf.py after writing the requested governor string to every core's scaling_governor sysfs file: it re-reads via GetCPUGovernor and the value still does not match, so the write did not take effect.

Source

Thrown at deps/v8/tools/run_perf.py:1147

      raise

  @staticmethod
  def SetCPUGovernor(value):
    try:
      cpu_indices = CustomMachineConfiguration.GetCPUCoresRange()
      for cpu_index in cpu_indices:
        cpu_device = CustomMachineConfiguration.GetCPUPathForId(cpu_index)
        with open(cpu_device, 'w') as f:
          f.write(value)

    except Exception:
      logging.exception('Failed to change CPU governor to %s. Are we '
                        'running under sudo?', value)
      raise

    cur_value = CustomMachineConfiguration.GetCPUGovernor()
    if cur_value != value:
      raise Exception('Could not set CPU governor. Present value is %s'
                      % cur_value )


class MaxTotalDurationReachedError(Exception):
  """Exception used to stop running tests when max total duration is reached."""
  pass


def Main(argv):
  parser = argparse.ArgumentParser(epilog="""example:
      ./run_perf.py --d8-path=out/Release/d8 $V8_PERF/benchmarks/JetStream/JetStream2.json
    """)
  parser.add_argument('--arch',
                      help='The architecture to run tests for. Pass "auto" '
                      'to auto-detect.', default='x64',
                      choices=SUPPORTED_ARCHS + ['auto'])
  parser.add_argument('--buildbot',
                      help='Deprecated',

View on GitHub (pinned to 1b2de5e052)

Solutions

1. Run the harness as root (sudo) so the sysfs write persists.
2. Confirm the requested governor is listed: cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors, and pick one that is present.
3. Stop conflicting daemons (thermald, TLP, power-profiles-daemon, cpufrequtils) that revert the governor, then re-run.
4. If the BIOS locks frequency control, enable OS frequency control in BIOS/UEFI.

Example fix

# before
./run_perf.py suite.json
# after
sudo systemctl stop thermald power-profiles-daemon
sudo cpupower frequency-set -g performance
sudo ./run_perf.py suite.json

def can_set_governor(value: str) -> bool:
    import glob, os
    avail = open('/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors').read().split()
    return value in avail and os.geteuid() == 0

null

try:
    CustomMachineConfiguration.SetCPUGovernor('performance')
except Exception as e:
    logging.warning('Could not pin CPU governor (%s); stop thermald/TLP/power-profiles-daemon and re-run with sudo.', e)

When it happens

Trigger: SetCPUGovernor(value) writes without IOError but GetCPUGovernor() returns a different value. Happens when not root (write appears to succeed but is discarded), when the requested governor is not in scaling_available_governors for the driver, when a BIOS/management engine or thermal daemon overrides the setting, or in a container with a read-only/virtualized sysfs.

Common situations: Running perf tests without sudo; requesting 'performance' when only 'powersave' is available on intel_pstate; a power daemon (thermald, TLP, power-profiles-daemon) re-applying its policy immediately; containerized runs.

Related errors

• CPU cores have differing governor settings
• Present value is %s
• --v8-enable-hugepage is supported only on linux.
• %s only allows TraceConfig and LeafTraceConfig as child conf
• Invalid suite configuration.%s

AI-assisted analysis of nodejs/node@1b2de5e052 (2026-08-13). Data as JSON: /api/errors/37ed0f9cbebc4c96. Report an issue: GitHub.