About transactions lost on data node failure - Colin Pitrat - com.mysql.lists.cluster

3 messages in com.mysql.lists.clusterAbout transactions lost on data node ...

From	Sent On	Attachments
Colin Pitrat	05 Dec 2007 07:34
Frazer Clement	05 Dec 2007 11:04
Frazer Clement	06 Dec 2007 00:20

Subject:	About transactions lost on data node failure
From:	Colin Pitrat (coli...@amadeus.com)
Date:	12/05/2007 07:34:38 AM
List:	com.mysql.lists.cluster

Hello,

I did a little benchmark of MySQL Cluster and it appeared that transaction are not continued when a data node crash. I found it really strange, as I thought that at least SQL nodes would handle this (I mean replay the transaction when it gets an error from the NDB cluster). Why isn't it done this way ? Would it be possible or is there something I miss that prevent SQL nodes from replaying the transaction ?