Revisions revert unexpectedly

Question

I have a really horrible intermittent bug where a collection of nodes will reset to a previous revision.

Here is an image of the revisions tab on the node:

There are three unusual aspects:

• The 'current revision' is at the bottom of the list and has lower vid than the rest.
• The date of the 'current revision' is yesterday but the low vid implies it was created a while ago.
• The revision log says "Updated by FeedsNodeProcessor" yesterday but the Feeds log tells me "Last run 1 week ago".

I'm hoping someone can point me in the right direction on where to look for a cause since I can't replicate the behaviour.

It feels like the vid in the {node} table has been changed. Along with the created date and revision timestamp. I'm really not sure how that could happen, though...

EDIT: Current Hypothesis

I have a custom module that updates a field. It uses a cron queue, so I'm wondering if the queue contains stale node data:

/**
 * Implements hook_cron().
 */
function ih_transcript_import_cron() {
  // Load all Video nodes.
  $videos = ih_transcript_import_load_video_nodes();

  $queue = DrupalQueue::get('ih_video_transcripts');
  foreach($videos as $video) {
    $queue->createItem($video);
  }
}

In my queue's 'worker callback', I should probably reload the node in case the object has been left in the queue for some time? And ensure I set $node->revision = TRUE before node_save?

Answer 1

I tested my hypothesis above and it was correct. The queue isn't fully processed during a cron run, so there are stale node objects sitting in the queue table in the database. The solution was to reload the node in the worker callback.

Original Code

/**
 * Implements hook_cron().
 */
function ih_transcript_import_cron() {
  // Load all Video nodes.
  $videos = ih_transcript_import_load_video_nodes();

  $queue = DrupalQueue::get('ih_video_transcripts');
  foreach($videos as $video) {
    $queue->createItem($video);
  }
}

/**
 * Implements hook_cron_queue_info().
 */
function ih_transcript_import_cron_queue_info() {
  $queues['ih_video_transcripts'] = array(
    'worker callback' => 'ih_transcript_import_import_transcript',
    'time' => '60',
  );
  return $queues;
}

/**
 * Implements callback_queue_worker().
 */
function ih_transcript_import_import_transcript($video_node) {
  // THIS IS THE PROBLEMATIC PART, $video_node is stale since it is
  // pulled from the queue table of the database and there may have been
  // multiple cron runs since it was created (2 days in my case).
  $video_node->field_transcript_text[$video_node->language][0]['value'] = "My value";
  node_save($video_node);
}

The Fix

/**
 * Implements callback_queue_worker().
 */
function ih_transcript_import_import_transcript($video_node) {
  // THIS IS THE IMPORTANT FIX, reload the video node to ensure
  // we are saving the most up to date data.
  $video_node = node_load($video_node->nid);

  $video_node->field_transcript_text[$video_node->language][0]['value'] = $transcript_text->data;
  node_save($video_node);
}

Better Fix

Save ONLY the node Id in the queue rather than the entire node. Also, I added some code to ensure that there are no duplicates in the queue. My queue table had 64000 records!!

/**
 * Implements hook_cron().
 */
function ih_transcript_import_cron() {
  // Load all Video nodes.
  $videos = ih_transcript_import_load_video_nodes();
  $currentQueueItems = array();
  $claimedItems = array();

  $queue = DrupalQueue::get('ih_video_transcripts');

  // We only want to add an item to a queue if it doesn't already exist in the
  // queue to avoid the queue table becoming massive.  Retrieve a list of
  // node ids that already exist in the queue.
  while($item = $queue->claimItem()) {
    $currentQueueItems[] = $item->data;

    // Store a copy of the claimed items, so we can release them again. We
    // only claim them in order to grab the node Id, so need to release again
    // to allow processing.
    $claimedItems[] = $item;
  }

  // Release claimed queue items to allow them to be processed by other workers.
  foreach($claimedItems as $claimedItem) {
    $queue->releaseItem($claimedItem);
  }

  foreach($videos as $video) {
    // Don't add videos that already exist in the queue.
    if(!in_array($video->nid, $currentQueueItems)) {
      $queue->createItem($video->nid);
    }
  }
}

/**
 * Implements hook_cron_queue_info().
 */
function ih_transcript_import_cron_queue_info() {
  $queues['ih_video_transcripts'] = array(
    'worker callback' => 'ih_transcript_import_import_transcript',
    'time' => '60',
  );
  return $queues;
}

/**
 * Implements callback_queue_worker().  Can also be called independently of cron
 * to import a single transcript.
 */
function ih_transcript_import_import_transcript($video_nid) {

  $video_node = node_load($video_nid);

  $video_node->field_transcript_text[$video_node->language][0]['value'] = $transcript_text->data;
  node_save($video_node);
}