stablyai/orca · error · AggregateError

Benchmark process group recovery failed before validation

Error message

Benchmark process group recovery failed before validation

What it means

In signalValidatedProcessGroup, the initial call to processIdentities(true) (which runs `ps eww -axo pid=,pgid=,command=`) throws an error before the group can even be validated. The recovery function compensateStoppedGroup (sending SIGCONT to stopped/anchor PIDs) also produces errors. Both the ps failure and recovery failures are wrapped in an AggregateError.

Source

Thrown at config/scripts/macos-computer-helper-owner-loss-processes.mjs:238

}

export function signalValidatedProcessGroup(
  pgid,
  environmentFragment,
  signal,
  groupState = { stopped: false, anchorPid: null },
  operations = processGroupSignalOperations
) {
  if (!Number.isInteger(pgid) || pgid <= 0) {
    return false
  }
  let members
  try {
    members = operations.processIdentities(true).filter((identity) => identity.pgid === pgid)
  } catch (error) {
    const recoveryErrors = compensateStoppedGroup(pgid, groupState, operations)
    if (recoveryErrors.length > 0) {
      throw new AggregateError(
        [error, ...recoveryErrors],
        'Benchmark process group recovery failed before validation'
      )
    }
    throw error
  }
  if (members.length === 0) {
    const recoveryErrors = compensateStoppedGroup(pgid, groupState, operations)
    if (recoveryErrors.length > 0) {
      throw new AggregateError(recoveryErrors, 'Benchmark missing process group recovery failed')
    }
    return false
  }
  if (members.some((identity) => !identity.command.includes(environmentFragment))) {
    const ownershipError = new Error('Benchmark process group no longer belongs to this trial')
    const recoveryErrors = compensateStoppedGroup(pgid, groupState, operations)
    if (recoveryErrors.length > 0) {
      throw new AggregateError(

View on GitHub (pinned to 1136503c6a)

Solutions

1. Check why `ps eww -axo pid=,pgid=,command=` failed — increase maxBuffer or reduce process count
2. Inspect AggregateError.errors: errors[0] is the ps failure, errors[1+] are recovery failures
3. Ensure the process group's stopped/anchor PIDs are still owned by the benchmark user
4. If ps output is too large, consider filtering to the target pgid earlier in the pipeline

Example fix

// before: ps eww may exceed buffer on systems with many processes
const output = execFileSync('ps', ['eww', '-axo', 'pid=,pgid=,command='], {
  encoding: 'utf8',
  maxBuffer: 20 * 1024 * 1024
})

// after: increase buffer or scope the query
const output = execFileSync('ps', ['eww', '-axo', 'pid=,pgid=,command='], {
  encoding: 'utf8',
  maxBuffer: 100 * 1024 * 1024
})

// Pre-check: ensure ps eww will not overflow
const psCount = execFileSync('ps', ['-axo', 'pid='], { encoding: 'utf8' }).trim().split('\n').length
if (psCount > 100_000) {
  throw new Error(`Process count (${psCount}) too high for ps eww — reduce before running`)
}

try {
  signalValidatedProcessGroup(pgid, envFragment, signal, groupState)
} catch (error) {
  if (error instanceof AggregateError) {
    // errors[0] = ps failure, errors[1+] = recovery failures
    const [psError, ...recoveryErrors] = error.errors
    console.error('ps eww failed:', psError.message)
    recoveryErrors.forEach(e => console.error('Recovery failed:', e.message))
  }
  throw error
}

When it happens

Trigger: processIdentities(true) throws (ps eww fails: output exceeds 20MB maxBuffer, EACCES, or ps binary not found) at line 234. Then compensateStoppedGroup also encounters non-ESRCH errors when trying to SIGCONT the stopped group or anchor PID. recoveryErrors.length > 0 at line 237.

Common situations: System with an extremely large number of processes causing ps eww output to exceed the 20MB buffer; permission restrictions on ps; the stopped/anchor PID was recycled to a process owned by another user (EPERM on SIGCONT); ps not in PATH.

Related errors

• Benchmark missing process group recovery failed
• Benchmark process group authority recovery failed
• Could not read watchdog child RSS for PID ${pid}
• Could not read CPU time for PID ${pid}
• Invalid process CPU time: ${value}

AI-assisted analysis of stablyai/orca@1136503c6a (2026-08-12). Data as JSON: /api/errors/02c14172ad0514c4. Report an issue: GitHub.