Customer stuck in an endless loop

Question

Problem: Some Customer (< 0.5%) got stuck in an endless loop if they visit our Shop (v1.9.2.4) or add a Product to cart. If the Customer delete the Cookies, everything works fine.

Because I can't reproduce the Problem on my local Maschine, I wrote a Script which do a curl Requests with every single session_id and log all 5xx errors.
With this Method I got a couple "bad" sessions and trace down a loop with xdebug. I figured out, that the endless loop appears if the flag sales_flat_quote.trigger_recollect is set to 1.

app/code/core/Mage/Sales/Model/Quote.php:1375

protected function _afterLoad()
{
    // collect totals and save me, if required
    if (1 == $this->getData('trigger_recollect')) {
        $this->collectTotals()->save();
    }
    return parent::_afterLoad();
}

Trace of the Loop:

1. The third-party Module Idev_OneStepCheckout listen to sales_quote_collect_totals_before and Call Mage::helper('payment')->getStoreMethods($store, $quote)
2. This Helper Call the method getConfigData() by a dynamic Model from the current Payment Methods. Module: Payone_Core
3. getConfig() call another Method which trigger the collectTotals()

A classic infinity Observer loop.

My first approach was to figure out where Magento set trigger_recollect to 1.

$ grep -nre "trigger_recollect\|triggerRecollect" --ignore-case

I found this two places:

app/code/core/Mage/Sales/Model/Resource/Quote.php:207 app/code/core/Mage/Sales/Model/Resource/Quote.php:282

Magento only call this Methods if you have Catalog Price Rules and change the affected Products. Then all quotes with this Product got the trigger_recollect flag. But we don't have any Price Rules and the Problem appears for new customer without a existing Quote, too.

Because of this two functions, I'm not longer sure if the specific loop is the result of the flag and this is very frustrating, because I can't explain why the trigger_recollect is set.

All logs (Magento, Nginx, PHP) are empty. Except this message:

Maximum execution time of 30 seconds exceeded

What I also tried was to compare the raw Data of a "normal" Session and a "bad" Session. Both are exactly the same.

A quick and dirty fix would be to rewrite two third party modules, and pass the quote through all method the prevent the getQuote() call, but there should be a reason why Magento set thetrigger_recollect flag or new customer git stuck, too.

Does anybody has a solution or an idea what i can also try or check?

Answer 1

looks like there are a lot more call of the particular methods which changes the trigger_recollect flag. The final solution was to pass the Quote through all methods. The Problem was the result of two third party modules.