When i want to distribute data along shards i specify the "routing" option in my Spark code (with Scala) like this example:
EsSparkSQL.saveToEs(df, conf.getString("s-app.persistenciaES"), Map("es.mapping.id" -> routing_field))
The spark process finish without any errors, but when i see my index stats, my index have documents on two columns: "docs.count" and "docs.deleted"
I dont have documentos on "docs.deleted" if i dont use routing option.
Do you know why some documents are distributed aleatory to "docs.deleted"?