E11000 duplicate key error index: MongoDb unusual error

Question

I have a simple "users" collection inside which right now i have only 2 documents.

{
    "_id": ObjectId("4ef8e1e41d41c87069000074"),
    "email_id": {
        "0": 109,
        "1": 101,
        "2": 64,
        "3": 97,

{
    "_id": ObjectId("4ef6d2641d41c83bdd000001"),
    "email_id": {
        "0": 109,
        "1": 97,
        "2": 105,
        "3": 108,

now if i try to create a new index with {unique: true} on email_id field, mongodb complaints me with "E11000 duplicate key error index: db.users.$email_id dup key: { : 46 }". I get same error even after specifying {dropDups: true}, however i don't think this is the case here, as both document have different email id's stored.

I am not sure what's going on here, any pointers will be greatly appreciated.

Edit: Full view of documents:

{
"_id": ObjectId("4ef8e1e41d41c87069000074"),
"email_id": {
 "0": 109,
 "1": 101,
 "2": 64,
 "3": 97,
 "4": 98,
 "5": 104,
 "6": 105,
 "7": 110,
 "8": 97,
 "9": 118,
 "10": 115,
 "11": 105,
 "12": 110,
 "13": 103,
 "14": 104,
 "15": 46,
 "16": 99,
 "17": 111,
 "18": 109
 }
}

and

{
"_id": ObjectId("4ef6d2641d41c83bdd000001"),
"email_id": {
 "0": 109,
 "1": 97,
 "2": 105,
 "3": 108,
 "4": 115,
 "5": 102,
 "6": 111,
 "7": 114,
 "8": 97,
 "9": 98,
 "10": 104,
 "11": 105,
 "12": 110,
 "13": 97,
 "14": 118,
 "15": 64,
 "16": 103,
 "17": 109,
 "18": 97,
 "19": 105,
 "20": 108,
 "21": 46,
 "22": 99,
 "23": 111,
 "24": 109
 }
}

There are a couple of more fields like "display_name", "registered_since", etc which i have omitted from the display above (i don't think they have any role in the error thrown, if you still need them i can probably paste the entire documents here)

I am using erlang mongodb driver for communication with my mongo instance. All fields as can be seen are saved as binary bytes, thats why you see such weird email_id in document.

Note: Binary byte format is not forced by my code logic, i very much pass string email_id inside my bson documents, but i always end up seeing my data as binary bytes. (Probably because how erlang mongodb driver is written, i didn't really investigate on this, since my find(), find_one() and other queries works as expected even with fields saved as binary bytes)

Edit: > db.users.findOne()

{
"_id" : ObjectId("4ef6d2641d41c83bdd000001"),
"email_id" : [
    109,
    97,
    105,
    108,
    115,
    102,
    111,
    114,
    97,
    98,
    104,
    105,
    110,
    97,
    118,
    64,
    103,
    109,
    97,
    105,
    108,
    46,
    99,
    111,
    109
],
"display_name" : [
    65,
    98,
    104,
    105,
    110,
    97,
    118,
    43,
    83,
    105,
    110,
    103,
    104
],
"provider" : [
    106,
    97,
    120,
    108,
    46,
    105,
    109
],
"provider_id" : [ ]
}

Answer 1

When MongoDB indexes an array field, it actually indexes the individual elements in the array. This is to efficiently support queries looking for a particular element of an array, like:

db.users.find({email_id: 46})

Since this email_id (46) exists in both documents, there are duplicate keys in your unique index.

I'm not sure why you would get this error if you have dropDups: true set... can you show a code sample with how you're invoking createIndex? You should also try dropDups: 1, as MongoDB erroneously treats 1 and true differently in this context (see https://jira.mongodb.org/browse/SERVER-4562).