OrientDB ETL, create edge between two vertices which are already in Graph

I am trying to create an edge between two vertices which are already part of OreintDB. My edge data is in a MySQL table.

Here is my oetl json.

    {
      "config": {
        "log": "info"
      },
      "source": { "file": { "path": "/Users/RP/user_invited_data.csv" } },
  "extractor": { "csv": {"columnsOnFirstLine": true, "columns":["user_id:string", "invited_by:string", "invited_date:datetime"] } },
      "transformers" : [
       { "vertex": { "class": "User", "skipDuplicates": true} },
        { "edge": { "class": "INVITED", "direction" : "in", 
                "joinFieldName": "invited_by",
                "lookup":"select expand(u) from (match {class: User, as: u} return u) where u.user_id = ?;", 
                "unresolvedLinkAction":"NOTHING",
                "edgeFields": { "invited_date": "${input.invited_date}" },
                "skipDuplicates": true
               } 
        },
        { "field": 
          { "fieldNames": 
            [ "invited_by", "invited_date"], 
            "operation": "remove"
          } 
      }
      ],
      "loader" : {
        "orientdb": {
          "dbURL": "remote:localhost/abcd_graph",
          "dbUser": "root",
          "dbPassword": "root",
          "dbType": "graph",
          "dbAutoCreate": false,
          "batchCommit": 1000
        }
      }
    }

When I run the above json, it is throwing ORecordDuplicatedException for the User vertex. I have a unique index created on user_id and have the skipDuplicates = true. Any suggestions would be greatly appreciated.

UPDATE: Gem of OrientDB, skipDuplicates actually works when your log level is not DEBUG. But the problem is not solved yet. No errors now but the edges are not created. I will keep debugging it and see if I can fix it tonight.

UPDATE After debugging a bit deeper, I got an exception deeper at the storage level.

com.orientechnologies.orient.core.exception.ODatabaseException: Impossible to serialize invalid link #-1:-1
    DB name="abcd_graph"
    at com.orientechnologies.orient.core.serialization.serializer.record.binary.ORecordSerializerBinaryV0.writeOptimizedLink(ORecordSerializerBinaryV0.java:867)
    at com.orientechnologies.orient.core.serialization.serializer.record.binary.ORecordSerializerBinaryV0.serializeValue(ORecordSerializerBinaryV0.java:754)
    at com.orientechnologies.orient.core.serialization.serializer.record.binary.ORecordSerializerBinaryV0.serialize(ORecordSerializerBinaryV0.java:385)
    at com.orientechnologies.orient.core.serialization.serializer.record.binary.ORecordSerializerBinary.toStream(ORecordSerializerBinary.java:99)
    at com.orientechnologies.orient.core.record.impl.ODocument.toStream(ODocument.java:2381)
    at com.orientechnologies.orient.core.record.impl.ODocument.toStream(ODocument.java:664)
    at com.orientechnologies.orient.core.db.document.ODatabaseDocumentTx.executeSaveRecord(ODatabaseDocumentTx.java:2183)
    at com.orientechnologies.orient.core.tx.OTransactionNoTx.saveRecord(OTransactionNoTx.java:191)
    at com.orientechnologies.orient.core.db.document.ODatabaseDocumentTx.save(ODatabaseDocumentTx.java:2758)
    at com.orientechnologies.orient.core.db.document.ODatabaseDocumentTx.save(ODatabaseDocumentTx.java:102)
    at com.orientechnologies.orient.core.record.impl.ODocument.save(ODocument.java:1805)
    at com.orientechnologies.orient.core.record.impl.ODocument.save(ODocument.java:1801)
    at com.tinkerpop.blueprints.impls.orient.OrientGraphNoTx.addEdgeInternal(OrientGraphNoTx.java:242)
    at com.tinkerpop.blueprints.impls.orient.OrientGraphNoTx.addEdgeInternal(OrientGraphNoTx.java:137)
    at com.tinkerpop.blueprints.impls.orient.OrientVertex.addEdge(OrientVertex.java:741)
    at com.tinkerpop.blueprints.impls.orient.OrientVertex.addEdge(OrientVertex.java:688)
    at com.orientechnologies.orient.etl.transformer.OEdgeTransformer.createEdge(OEdgeTransformer.java:203)
    at com.orientechnologies.orient.etl.transformer.OEdgeTransformer.executeTransform(OEdgeTransformer.java:123)
    at com.orientechnologies.orient.etl.transformer.OAbstractTransformer.transform(OAbstractTransformer.java:39)
    at com.orientechnologies.orient.etl.OETLPipeline.execute(OETLPipeline.java:110)
    at com.orientechnologies.orient.etl.OETLProcessor$OETLPipelineWorker.call(OETLProcessor.java:620)
    at com.orientechnologies.orient.etl.OETLProcessor$OETLPipelineWorker.call(OETLProcessor.java:601)
    at java.util.concurrent.FutureTask.run(FutureTask.java:266)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)

UPDATE I have changed the extractor from DB to CSV, so that it will be easier to reproduce.

Create Schema:

CREATE class User IF NOT EXISTS extends V;
create property User.user_id IF NOT EXISTS String;
create property User.name IF NOT EXISTS String;
create index user_idx on User(user_id) unique;

insert into User set user_id = '1000_USER1', name = 'Bob';
insert into User set user_id = '1001_USER2', name = 'Robert';

Sample CSV:

user_id, ivited_by, invited_date
1001_USER2, 1000_USER1,

After some struggle and reread the whole ETL documentation and some debugging, I figured it out.

We need to use MERG transformer instead of VERTEX. Merge transformer will lookup for a Vertex instead of creating it.

Here is my json looks like

"transformers" : [
    { "merge": { "joinFieldName": "user_id", "lookup": "User.user_id" } },
    { "edge": { "class": "INVITED", "direction" : "out", 
            "joinFieldName": "invited_by",
            "lookup": "SELECT expand(u) from (match {class: User, as: u} return u) where u.user_id = ?", 
            "unresolvedLinkAction":"NOTHING",
            "edgeFields": { "invited_date": "${input.invited_date}" },
            "skipDuplicates": true
           } 
    },
    { "field": 
      { "fieldNames": 
        [ "invited_by", "invited_date"], 
        "operation": "remove"
      } 
  }
  ]

I still have one another problem, but I will take it as a separate thing and research about it. The problem is it is creating duplicate edges between the same two vertices

I will tackle it as a separate problem.

One thing I observed all the times with OrientDB is, the stuff is there, but it is hard to figure it out.