Database – Rebuild Index Taking Forever in Magento

There was a presentation from Maksym Iaroshenko at Developers Paradise – Friends, Fun, Snow: Magento non-MySQL implementations

I also found this video (same guy) presenting at Meet Magento Poland: "Magento non-MySQL implementations" Maksym Iaroshenko, Dmytro Vasylenko, Eltrino

During one of the Magento Hackathons this repo was created: MongoDB-OrderTransactions

I don't see a working version of Magento 2 on anything else than MySQL in the next 12 months (at least).

The challange with another RDBMS and Magento is that just porting the database tables with the same schema won't be a solution, because some refactoring is needed, some stuff should be moved at a database level (depends on RDBMS, would be nice to see stored views used).

Another challange is providing a set of tools for the actual migration and management of the environment so developers that would want to contribute not to break their heads for couple of weeks just to get some kind of functional dev environment.

Start a repo on github and see how other people feel about this.

The catalog_product_index_eav table has a foreign key contraint named

FK_CAT_PRD_IDX_EAV_ENTT_ID_CAT_PRD_ENTT_ENTT_ID

Looking at this table's definition

CREATE TABLE `catalog_product_index_eav` (
  `entity_id` int(10) unsigned NOT NULL DEFAULT '0' COMMENT 'Entity ID',
  `attribute_id` smallint(5) unsigned NOT NULL DEFAULT '0' COMMENT 'Attribute ID',
  `store_id` smallint(5) unsigned NOT NULL DEFAULT '0' COMMENT 'Store ID',
  `value` int(10) unsigned NOT NULL DEFAULT '0' COMMENT 'Value',
  PRIMARY KEY (`entity_id`,`attribute_id`,`store_id`,`value`),
  KEY `IDX_CATALOG_PRODUCT_INDEX_EAV_ENTITY_ID` (`entity_id`),
  KEY `IDX_CATALOG_PRODUCT_INDEX_EAV_ATTRIBUTE_ID` (`attribute_id`),
  KEY `IDX_CATALOG_PRODUCT_INDEX_EAV_STORE_ID` (`store_id`),
  KEY `IDX_CATALOG_PRODUCT_INDEX_EAV_VALUE` (`value`),
  CONSTRAINT `FK_CATALOG_PRODUCT_INDEX_EAV_STORE_ID_CORE_STORE_STORE_ID` FOREIGN KEY (`store_id`) REFERENCES `core_store` (`store_id`) ON DELETE CASCADE ON UPDATE CASCADE,
  CONSTRAINT `FK_CAT_PRD_IDX_EAV_ATTR_ID_EAV_ATTR_ATTR_ID` FOREIGN KEY (`attribute_id`) REFERENCES `eav_attribute` (`attribute_id`) ON DELETE CASCADE ON UPDATE CASCADE,
  CONSTRAINT `FK_CAT_PRD_IDX_EAV_ENTT_ID_CAT_PRD_ENTT_ENTT_ID` FOREIGN KEY (`entity_id`) REFERENCES `catalog_product_entity` (`entity_id`) ON DELETE CASCADE ON UPDATE CASCADE
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COMMENT='Catalog Product EAV Index Table';

we can see the foreign key definition is

CONSTRAINT `FK_CAT_PRD_IDX_EAV_ENTT_ID_CAT_PRD_ENTT_ENTT_ID` 
FOREIGN KEY (`entity_id`) 
REFERENCES `catalog_product_entity` (`entity_id`) 
ON DELETE CASCADE ON UPDATE CASCADE

This means for every entity_id row in catalog_product_index_eav, there needs to be an identical, corresponding entity_id value in catalog_product_entity.

The root of your problem is for some reason (either a rouge extension, errors caused by randomly typing in SQL from the internet, or performing data updates with disabled indexes), Magento's indexing attempts to update data in catalog_product_index_eav that violates this rule. The next step is identifying what Magento's doing so you can fix the data.

If we look at your call stack, this looks like a good place to start debugging

#10 app/code/core/Mage/Catalog/Model/Resource/Product/Indexer/Eav/Abstract.php(54): 
Mage_Index_Model_Resource_Abstract->syncData()

Jumping to that source file, we see the following bit of code

public function syncData()
{
    $this->beginTransaction();
    try {
        /**
         * Can't use truncate because of transaction
         */
        $this->_getWriteAdapter()->delete($this->getMainTable());
        $this->insertFromTable($this->getIdxTable(), $this->getMainTable(), false);
        $this->commit();
    } catch (Exception $e) {
        $this->rollBack();
        throw $e;
    }
    return $this;
}

As part of its indexing process, Magento tries to sync data from an "index table" (getIdxTable), to a "source" (getMainTable) table.

public function insertFromTable($sourceTable, $destTable, $readToIndex = true)
{
    //...
}

For this particular index, the index table is catalog_product_index_eav_idx, and the source table is catalog_product_index_eav.

Note: Be careful with your terminology around here, things are confusingly named. The "Source" table is the table we're copying to. (I believe it's called the source table because it's the "source" a normal Magento system will query from when it needs information)

So, Magento is trying to sync a row from catalog_product_index_eav_idx to the table catalog_product_index_eav. However, this causes the previously mentioned foreign key error. This leads us to two possible conclusions

1. The catalog_product_index_eav_idx has entity_id rows that do not exist in catalog_product_entity.

2. The catalog_product_index_eav table has (through previous manipulation with index checks turned off) entity_id rows that do not exist in catalog_product_entity.

So, your mission here is to figure out which entity_id rows in catalog_product_index_eav and catalog_product_index_eav_idx don't exist in catalog_product_entity, and manually delete said rows (from catalog_product_index_eav and catalog_product_index_eav_idx).

If it were me, and my catalog_product_entity table wasn't too large, I'd start with the following queries (these are untested, as I don't have any Magento tables with the above invalid data states)

SELECT * 
FROM catalog_product_index_eav_idx 
WHERE NOT (entity_id IN (SELECT entity_id from catalog_product_entity));

SELECT * FROM catalog_product_index_eav 
WHERE NOT (entity_id IN (SELECT entity_id from catalog_product_entity));

Good luck!