Hi Manuel, Thanks a lot for reporting this!
Yes, this issue is most likely related to the recent changes to shading the Elasticsearch connector dependencies, though it is a bit curious why I didn’t bump into it before while testing it. The Flink job runs Lucene queries on a data stream which ends up in an Elasticsearch index. Could you explain a bit more where the Lucene queries are executed? Were there other dependencies required for this? I would highly appreciate any opinion on this workaround. Could it have side effect ? I think your workaround wouldn’t be harmful. Could you explain how you came to the solution? That would help me in getting to the bottom of the problem (and maybe other potential similar issues). Cheers, Gordon On 23 March 2018 at 12:43:31 AM, Till Rohrmann (till.rohrm...@gmail.com) wrote: Hi Manuel, thanks for reporting this issue. It sounds to me like a bug we should fix. I've pulled Gordon into the conversation since he will most likely know more about the ElasticSearch connector shading. Cheers, Till On Thu, Mar 22, 2018 at 5:09 PM, Haddadi Manuel <manuel.hadd...@gfi.fr> wrote: Hello, When upgrading from flink-1.3.2 to flink-1.4.2, I faced this error on runtime of a Flink job : java.util.ServiceConfigurationError: An SPI class of type org.apache.lucene.codecs.PostingsFormat with classname org.apache.lucene.search.suggest.document.Completion50PostingsFormat does not exist, please fix the file 'META-INF/services/org.apache.lucene.codecs.PostingsFormat' in your classpath. I added lucene-suggest dependency and then I encountered this : java.lang.ClassCastException: class org.elasticsearch.search.suggest.completion2x.Completion090PostingsFormat The Flink job runs Lucene queries on a data stream which ends up in an Elasticsearch index. It seems to me that this exception is a side effect of shading flink-connector-elasticsearch-5 dependencies. Actually, the only solution I have found is to re-build flink-connector-elasticsearch-5 jar excluding META-INF/services/org.apache.lucene.codecs.* I would highly appreciate any opinion on this workaround. Could it have side effect ? Thanks. And by the way, congrats to all Flink contributors, this is a pretty good piece of technology ! Regards, Manuel Haddadi
