To　improve　the　stability　and　reliability　of　the　ETL　workflow　engine,　in　this　paper,　we　put　forward　a　new　MAS-based　and　fault-tolerant　distributed　ELT　workflow　engine　which　is　used　to　study　the　log　management　and　exception　handling　mechanism　in　the　distributed　computing　environment.　In　the　new　engine,　the　ETL　jobs　will　be　split　into　several　job　domains　firstly.　Secondly,　when　an　error　happens　in　a　job　domain,　the　engine　will　catch　the　exception　from　the　error　logs　and　then　invoke　our　new　recovery　algorithm,　Batch,　to　recover　the　job　domain.　The　experiment　result　shows　that　our　engine　is　not　only　much　more　stabile　and　reliable　but　also　more　efficiency　in　the　execution　of　ETL　workflow.　©　2012　IEEE.