_mget and _search differences on ElasticSearch - elasticsearch

Question

I've indexed 2 documents:
As you can see, after having indexed those ones, I can see them in a search result:
[root#centos7 ~]# curl 'http://ESNode01:9201/living/fuas/_search?pretty'
{
"took" : 20,
"timed_out" : false,
"_shards" : {
"total" : 5,
"successful" : 5,
"failed" : 0
},
"hits" : {
"total" : 2, <<<<<<<<<<<<<<<<
"max_score" : 1.0,
"hits" : [ {
"_index" : "living",
"_type" : "fuas",
"_id" : "idFuaMerge1", <<<<<<<<<<<<<<<
"_score" : 1.0,
"_source":{"timestamp":"2015-10-14T16:13:49.004Z","matter":"null","comment":"null","status":"open","backlogStatus":"unknown","metainfos":[],"resources":[{"resourceId":"idResourceMerge1","noteId":"null"},{"resourceId":"idResourceMerge2","noteId":null}]}
}, {
"_index" : "living",
"_type" : "fuas",
"_id" : "idFuaMerge2", <<<<<<<<<<<<<<<<<<
"_score" : 1.0,
"_source":{"timestamp":"2015-10-14T16:13:49.004Z","matter":"null","comment":"null","status":"open","backlogStatus":"unknown","metainfos":[],"resources":[{"resourceId":"idResourceMerge3","noteId":null}]}
} ]
}
}
After that, I perform a multiget request setting the document ids:
[root#centos7 ~]# curl 'http://ESNode01:9201/living/fuas/_mget?pretty' -d '
{
"ids": ["idFuaMerge1", "idFuaMerge2"]
}
'
{
"docs" : [ {
"_index" : "living",
"_type" : "fuas",
"_id" : "idFuaMerge1",
"found" : false <<<<<<<<<<<<<<<<<<<<!!!!!!!!!!!!!!
}, {
"_index" : "living",
"_type" : "fuas",
"_id" : "idFuaMerge2",
"_version" : 4,
"found" : true, <<<<<<<<<<<<<<<!!!!!!!!!!!!!!!!!
"_source":{"timestamp":"2015-10-14T16:13:49.004Z","matter":"null","comment":"null","status":"open","backlogStatus":"unknown","metainfos":[],"resources":[{"resourceId":"idResourceMerge3","noteId":null}]}
} ]
}
How on earth, on a multiget request, the first document is NOT found and the other one does?

Answer 1

This can only happen if you have used routing key to index your document. Or even parent child relation can also imply the same.
When a document is given for indexing , that document is mapped to a unique shard using the mechanism of routing. In this mechanism the docID is converted to a hash and modulas operation of that hash is taken to determine to which shard the document should go.
So in short
for documentA by default the shard might be 1. Default shard is computed based on routing key.
But then because you applied the routing key yourself , this document is mapped to a different shard , tell 0.
Now when you try to get the document without the routing key , it expects the document to be in shard 1 and not shard 0 and hence your multi get fails as it directly looks in shard 1 to get the document.
The search works because search operation happens across all shards/