Multithreaded XML Import (San Francisco Magento Meetup)

Author: Fabrizio Branca

1. Multithreaded XML Import …for Magento San Francisco Magento Meetup Group - October 23, 2013
2. Fabrizio Branca Lead System Developer at
3. E-Commerce: Magento CMS: TYPO3 Global Enterprise Projects Portals: ZF, FLOW,… High Performance /Scale Mobile Searchperience: SOLR 120 people in 7 offices world-wide
4. Aoe_Import github.com/AOEmedia/Aoe_Import git clone --recursive …
5. Will Aoe_Import be the fastest product importer around? YES, of course! Well, maybe… Actually, Aoe_Import is only a XML Importer “Framework”. It’s up to you to decide how to handle the xml snippets…
6. for large XML files XML! Not CSV. full flexibility in processor implementation Aoe_Import multi-thread support! Subscribe your “Processors” to xpaths Stream processing (XMLReader) “event” driven
7. memory single product Problem Memory limit time
8. Memory limit time memory memory Trivial Solution Memory limit time
9. Beat the memory Leak by forking Waiting for other thread to terminate Threading overhead Process import
10. Forking? In PHP? $pid = pcntl_fork(); if ($pid) { // parent process runs what is here echo "parentn"; } else { // child process runs what is here echo "childn"; }
11. Threadi github.com/AOEmedia/Threadi
12. Clean OOP interface for PHP to forking and process management Threadi
13. Batch Processor Collect a bunch of imports … …fork… …and process them in a child process.
14. No imports are processed in the main thread. So there’s no memory leak happing here Main thread memory Memory limit time Create process collection Waiting for other thread to terminate Threading Process imports in process collection overhead Forks Every fork starts with the low memory footprint of the main thread Find the number of imports that can be processed at a time without hitting the memory limit
15. Multi-threading? Sure! Number of threads processed in parallel Number of items in a batch
16. Problems? Database Connection Database connection doesn’t like to be cloned! Mage::getSingleton('core/resource') ->getConnection('core_write') ->closeConnection();
17. Problems? Thread Safety
18. Problems? Thread Safety --- a/app/code/core/Enterprise/Catalog/Model/Index/Action/Catalog/Category/Product/Refresh.php +++ b/app/code/core/Enterprise/Catalog/Model/Index/Action/Catalog/Category/Product/Refresh.php @@ -326,7 +326,7 @@ class Enterprise_Catalog_Model_Index_Action_Catalog_Category_Product_Refresh ->setComment('Catalog Category Product Index Tmp'); $this->_connection->dropTable($this->_getMainTmpTable()); $this->_connection->createTable($table); $this->_connection->createTemporaryTable($table); + } /**
19. Other Use-Cases? Scheduler Queue processing Indexes Everything that’s batchable
20. Thank you! Any questions? My blog http://www.aoemedia.com http://www.fabrizio-branca.de @fbrnc Follow me on twitter!