Hi, this turned out to be a version conflict. The cluster was running on the latest 0.7 snapshot, and my code ran on the 0.7 release. Even when I used the 0.7 snapshot in my code again, it did not work. We updated the cluster to the release now as well, and it works again.
For the future: do they need to be at the exact same version (i.e. commit) or only when there are serious changes in-between? Thanks for pointing me in the right direction. On Tue, Nov 11, 2014 at 11:17 AM, Stefan Bunk <[email protected]> wrote: > Hi, > > I am using Flink 0.7 both in my program and on the cluster. > > Cheers > Stefan > > On Tue, Nov 11, 2014 at 7:48 AM, Aljoscha Krettek <[email protected]> > wrote: > >> Hi, >> which version of Flink are you using? >> >> Cheers, >> Aljoscha >> >> On Tue, Nov 11, 2014 at 12:09 AM, Stefan Bunk >> <[email protected]> wrote: >> > Hi Flinkers, >> > >> > I get the following error right after starting/deploying my program on >> the >> > cluster: >> > java.lang.RuntimeException: The initialization of the DataSource's >> outputs >> > caused an error: Could not load deserializer from the configuration. >> > ... >> > Caused by: java.io.InvalidClassException: >> > org.apache.flink.api.scala.typeutils.CaseClassSerializer; local class >> > incompatible: stream classdesc serialVersionUID = 7795994801161867172, >> local >> > class serialVersionUID = 1572449491425739294 >> > >> > I see this is an serialization issue, but I do not get why. >> > My workflow: >> > I call `mvn package` to build a jar with all dependencies, then execute >> my >> > program with the remote environment, referencing my newly created *.jar. >> > >> > What exactly does that tell me, and where do I have to look? >> > >> > Cheers >> > Stefan >> > >
