Mongo gives duplicate key error on _id_ field in Meteor application

Question

I'm searching a long time and a lot of topics about this issue. Until now I couldn't find any solution. Moreover, it's not all clear to me, hopefully you can help. Here's my question:

I designed a Meteor application and there is a collection in the Mongo DB with orders. That collection is filled by reading a csv file

import_file_orders = function(file) {
var lines = file.split('%\r\n');
var l = lines.length - 1;
for (var i=0; i < l; i++) {
  var line = lines[i];
  var line_parts = line.split('|');
  var ex_key = line_parts[0];
  var ex_name = line_parts[1];
  var clin_info = line_parts[2];
  var order_info = line_parts[3];
  var clinician_last_name = line_parts[4];
  var clinician_first_name = line_parts[5];
  var clinician_code = line_parts[6];
  var clinician_riziv = line_parts[7]
  var pat_id = line_parts[8];
  Meteor.orders.insert({Patient:pat_id, Exam_code:ex_key, Exam_name:ex_name, Clinical_info:clin_info, Order_info:order_info, Clinician:{first:clinician_first_name, last:clinician_last_name, c_code:clinician_code, riziv:clinician_riziv}, Planned:null});
  console.log("%");
};
}

After reading the CSV file there are errors for some documents in the collection:

duplicate key error index: protocolplanner.Orders.$_id_ dup key: { :     "2ZGvRfuD8iMvRiXJd" } insert failed

When I run the Mongo command db.Orders.getIndexes() I see that there are two indexes:

{
  "v" : 1,
  "key" : {
           "_id" : 1
   },
   "name" : "_id_",
   "ns" : "protocolplanner.Orders"
 }

It seems that there are two indexes: one _id index (which is always there and can't be deleted) and one _ id _ index. It seems that the _ id _ index caused the error. So I have three questions:

First: Why is there an _ id _ index? I defined no indexes in my Meteor code. Second: Why is there a dup key error for that index? Third: Also it seems that I can't remove the _ id _ index. Why is it? I now that you can't remove a _id index but in my opinion this isn't an _id index.

As you can see I'm totally lost. Please help!

EDIT:

As commented below, a little bit more info:

The amount of data I'm reading is 10151 lines. The function that reads the file in defined client side. Via allow and deny rules only admin users can enter the data into mongo. The lines are read correctly. After reading the file all data is available in the app. After a few seconds the index is automatically created by Mongo and the error appear. From then the lines for which the error occurs aren't visible anymore in the app.

I tried the following in the Mongo shell: db.Orders.find({_id:"2ZGvRfuD8iMvRiXJd"})

Mongo gives me the right document. This proves that the _id indeed is created by Meteor when the data is inserted in the DB. However this _id should be unique so I'm totally confused about the error I have.

EDIT 2: After some trial and error I have some new information about this problem. Maybe It's interesting to know, so we can find an answer on this problem.

Like described above, when I read the data on client side, I have the duplicate key error, even when I use ObjecID instead of Meteor ID. However when I push the data directly into Mongo via mongoinsert command, all the data is imported well and no error occurs. It seems that there's a conflict between server and client when I insert this amount of data (maybe asynchronous timing issues).

At this moment I'm searching a solution to read the data server side in the hope no error occurs.

Answer 1

So far it's not clear why there is the problem of the duplicate key error. However I tried some things and I found a workable solution.

I moved the insert of the data from client to server side. Therefore I followed the solution in this topic:

How to import data from CSV file into Meteor collection at server side

When the insert function is server side the duplicate key error doesn't appear and everything works perfectly.

Answer 2

This won't solve your problem, but it should point you in the right direction and maybe enable you to isolate the problem that you can use to create a new question:

First: Why is there an _id_ index?

There isn't. There's only one index, and it has a name and a key descriptor. That's not the same thing. The name of the default index is _id_, its key is _id.

Why is there a dup key error for that index?

The _id is normally created client-side, not server-side. The question is where those keys come from, because 2ZGvRfuD8iMvRiXJd is certainly not an ObjectId. This might be a meteor key, or you use some custom primary key, but I don't know how these keys are generated. Maybe whatever generates the key is susceptible to collisions?

More info on that would be helpful, but I'd suggest phrasing a new question so the question doesn't grow too large or get a lot of history.

Third: Also it seems that I can't remove the _ id _ index. Why is it?

That's a lemma of the first answer: You can't delete the mandatory primary key index.

Edit:

Meteor, by default, generates ids in a different way than MongoDb. That makes sense, because the convention for ObjectId makes collisions probable if the number of clients is large (i.e. if the clients aren't server instances, but client browsers, of which there are probably 2-3 orders of magnitude more).

Instead, Meteor apparently uses method to consistently generate pseudo-random numbers on client and server. Irritatingly, the implementation uses a PRNG and falls back to a not crypto-strong deterministic random number generator (Alea). In other words, finding out how exactly your ids are being generated could be tricky because it depends on a lot of details of your environment.

Workaround: Try to use ObjectId as a primary key:

Orders= new Meteor.Collection('Orders', {idGeneration: 'MONGO'});