phacility/phabricator · error · Exception

Patch generation took longer than configured limit ("%s") of

Error message

Patch generation took longer than configured limit ("%s") of %s second(s).

What it means

While publishing commit notifications, loadRawPatchText() calls diffusion.rawdiffquery with a timeout taken from the metamta.diffusion.time-limit config (PhabricatorRepositoryCommitPublishWorker.php:377-400). When generating the raw diff exceeds that many seconds, the conduit response sets tooSlow and the worker throws; the publish task then retries and keeps failing until the limit is raised or the diff becomes cheaper to compute.

Source

Thrown at src/applications/repository/worker/PhabricatorRepositoryCommitPublishWorker.php:394

    $time_key = 'metamta.diffusion.time-limit';
    $byte_key = 'metamta.diffusion.byte-limit';
    $time_limit = PhabricatorEnv::getEnvConfig($time_key);
    $byte_limit = PhabricatorEnv::getEnvConfig($byte_key);

    $diff_info = DiffusionQuery::callConduitWithDiffusionRequest(
      $viewer,
      $drequest,
      'diffusion.rawdiffquery',
      array(
        'commit' => $identifier,
        'linesOfContext' => 3,
        'timeout' => $time_limit,
        'byteLimit' => $byte_limit,
      ));

    if ($diff_info['tooSlow']) {
      throw new Exception(
        pht(
          'Patch generation took longer than configured limit ("%s") of '.
          '%s second(s).',
          $time_key,
          new PhutilNumber($time_limit)));
    }

    if ($diff_info['tooHuge']) {
      $pretty_limit = phutil_format_bytes($byte_limit);
      throw new Exception(
        pht(
          'Patch size exceeds configured byte size limit ("%s") of %s.',
          $byte_key,
          $pretty_limit));
    }

    $file_phid = $diff_info['filePHID'];
    $file = id(new PhabricatorFileQuery())

View on GitHub (pinned to 5720a38cfe)

Solutions

Raise the limit: bin/config set metamta.diffusion.time-limit 120 (or whatever fits your largest commits)
Improve generation speed: faster storage for the repository working copies, reduce load on the VCS host
Re-run publishing for stuck commits after raising the limit: bin/repository reparse --message <monogram>

Example fix

# before
bin/config set metamta.diffusion.time-limit 30

# after
bin/config set metamta.diffusion.time-limit 300

Defensive patterns

Strategy: retry

Validate before calling

// Size the timeout before publishing: probe the repo's worst-case commit.
// Raise the config to cover it:
//   bin/config set metamta.diffusion.time-limit 300
// Larger than the slowest 'git show'/'svn diff' of your biggest commit.

Try / catch

try {
  $patch = $this->loadRawPatchText($repository, $commit);
} catch (Exception $ex) {
  // tooSlow is transient for the task: after raising
  // metamta.diffusion.time-limit, let the worker retry (it will), or
  // manually requeue: bin/repository reparse --message <monogram>
  phlog($ex);
  throw $ex;
}

Prevention

Set metamta.diffusion.time-limit above the slowest diff generation in your history (measure with git show on the largest commit)
Keep repository working copies on fast local storage
Watch the publish queue for repeated tooSlow failures after big merges

When it happens

Trigger: A commit whose raw diff (3 lines of context) takes longer than metamta.diffusion.time-limit seconds to generate: giant merges, huge tree rewrites, binary-heavy commits, or mirrors on slow network storage.

Common situations: Monorepos with enormous commits; default limits too tight for the repository's normal history; initial imports flooding the publish queue; slow I/O on the daemon host.

Related errors

AI-assisted analysis of phacility/phabricator@5720a38cfe (2026-08-21). Data as JSON: /api/errors/9b9fac308add31bb. Report an issue: GitHub.