I recently had an issue with a large feed (~400 items) which is creating nodes and running through some external processing (calais). It usually works well but I think it encountered a timeout or network error during the last cron refresh which I'm guessing may have been during node_save. The result was that feedapi thought it had processed the feed and nothing happened on manual refresh. The only way to refresh was to manually reset the hash field in the db (I tried most of the other fields). This might be treating the symptoms but I'm wondering if it's worth having some kind of admin-only 'hard reset' link for each feed which will clear the hash and last updated fields.

Comments

alex_b’s picture

alex_b commented
Category: feature » bug

This is actually a bug. If feedapi does not finish, it should not set the hash. Can you confirm this behavior for the latest version? I'm pretty sure you will.

adub’s picture

adub commented

I think it will be pretty difficult for me a) to change version (as this is on a tightly controlled environment) b) repeat the error. I could scan the code and look for a suspect area if that helps but I can't really test this unfortunately.

aron novak’s picture

aron novak
Primary language Hungarian
Location Hungary, Budapest
commented 
// Closing step: Call after refresh and update feed statistics
  foreach (module_implements('feedapi_after_refresh') as $module) {
    $func = $module .'_feedapi_after_refresh';
    $func($feed);
  }

  // Set next_refresh_time to FEEDAPI_CRON_NEVER_REFRESH if refresh_time is FEEDAPI_CRON_NEVER_REFRESH.
  $next_refresh_time = $settings['refresh_time'] == FEEDAPI_CRON_NEVER_REFRESH ? $settings['refresh_time'] : (time() + $settings['refresh_time']);
  db_query("UPDATE {feedapi} SET next_refresh_time = %d, half_done = %d, hash = '%s' WHERE nid = %d", $next_refresh_time, $half_done, $feed->hash, $feed->nid);

Of course FeedAPI alters hash at the very last step of this process.