Problem/Motivation

Submissions exports have the following format 'dir/webfrom/[webform_id].ext or ..., the filename is the save on every export.
\Drupal\webform\Plugin\WebformExporterBase::getBaseFileName

  /**
   * {@inheritdoc}
   */
  public function getBaseFileName() {
    $webform = $this->getWebform();
    $source_entity = $this->getSourceEntity();
    if ($source_entity) {
      return $webform->id() . '.' . $source_entity->getEntityTypeId() . '.' . $source_entity->id();
    }
    else {
      return $webform->id();
    }
  }

On sites with several web nodes sharing the export directory over NFS, each batch request can hit a different node. When one node deletes and recreates the file, another node may still hold the old file handle in its NFS client cache, so `fopen()` fails:


Warning: fopen(/app/tmp/webform/my_form.csv): Failed to open stream: Stale file handle in
Drupal\webform\Plugin\WebformExporter\TabularBaseWebformExporter->createExport()
TypeError: fputcsv(): Argument #1 ($stream) must be of type resource, false given in fputcsv()
(line 116 of src/Plugin/WebformExporter/DelimitedWebformExporter.php).

The shared file name also makes two concurrent exports of the same webform write to (and delete) the same file.

Steps to reproduce

  • Run Drupal on 2+ web nodes with the Webform export temp directory on a shared NFS mount.
  • Export the results of a webform with enough submissions to require a batch, then download the file.
  • Export the same webform again right away. Some batch requests fail with the error above on the node that cached the deleted file.

Proposed resolution

Use a token for each batch to avoid filename collisions between batch executions.

Remaining tasks

TBD.

User interface changes

TBD.

API changes

TBD.

Data model changes

TBD.

Issue fork webform-3632332

Command icon Show commands

Start within a Git clone of the project using the version control instructions.

Or, if you do not have SSH keys set up on git.drupalcode.org:

About issue forks

Comments

eduardo morales alberti created an issue. See original summary.

eduardo morales alberti opened merge request !963

eduardo morales alberti’s picture

eduardo morales alberti
Primary language Spanish
Location Spain, 🇪🇺
commented
Status: Active » Needs review

Created MR:

Generate a token prefix:

$export_options['export_token'] = bin2hex(random_bytes(8));

The token is only internal; it will never download the file with it:

    if ($this->submissionExporter->getExporter()) {
      $base_file_name = $this->submissionExporter->getExporter()->getBaseFileName();
      $file_name = preg_replace('#^(' . preg_quote($base_file_name, '#') . ')\.[a-f0-9]{16}\.#', '$1.', basename($file_path));
      $response->setContentDisposition($download ? 'attachment' : 'inline', $file_name);
    }

Added a custom method openFileHandle, to catch the file when it is not available

 protected function openFileHandle($mode) {
    $file_path = $this->getExportFilePath();
    $file_handle = fopen($file_path, $mode);
    if ($file_handle === FALSE) {
      throw new \RuntimeException(sprintf('Unable to open the export file %s.', $file_path));
    }
    return $file_handle;
  }

Delete orphan export files using cron:

    $directory = \Drupal::service('webform_submission.exporter')->getFileTempDirectory();
    $expire = \Drupal::time()->getRequestTime() - 86400;
    foreach (scandir($directory) ?: [] as $file_name) {
      $file_path = $directory . '/' . $file_name;
      if (preg_match('/\.[a-f0-9]{16}\.(tar\.gz|[a-z0-9]+)$/', $file_name) && is_file($file_path) && filemtime($file_path) < $expire) {
        @unlink($file_path);
      }
    }