Hi Ionut, Yes, this is correct. Transaction is thread-local and I don't think what you propose is feasible, because someone has to coordinate the transaction. Map-reduce approach works for data processing, but not for transactional updates.
I understand that you don't know the keys in advance, but you can load them using SQL query like 'select _key from ... where ...' which will return only keys, not values. You can then update them using entry processor in a single transaction. As for the invocation on the client, this is a know issue: https://issues.apache.org/jira/browse/IGNITE-3471. I should be fixed soon. -Val -- View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Update-performance-tp6214p6284.html Sent from the Apache Ignite Users mailing list archive at Nabble.com.
