SERVER-24883 cloner should not create _id index when original collection doesn't have one

2 files changed, 32 insertions, 1 deletions

diff --git a/jstests/core/copydatabase_no_id_index.js b/jstests/core/copydatabase_no_id_index.js
new file mode 100644
index 00000000000..2cd1225d78b
--- /dev/null
+++ b/jstests/core/copydatabase_no_id_index.js
@@ -0,0 +1,23 @@
+/**
+ * This test creates a collection with autoIndexId: false. It then copies the database and expects
+ * the new collection to not have an _id index either.
+ */
+
+(function() {
+    "use strict";
+    var db1 = db.getSiblingDB('copydatabase_no_id_index');
+    var db2 = db.getSiblingDB('copydatabase_no_id_index2');
+    db1.dropDatabase();
+    db2.dropDatabase();
+
+    assert.commandWorked(db1.runCommand({create: 'foo', autoIndexId: false}));
+    assert.writeOK(db1.foo.insert({a: 1}));
+    assert.eq(db1.foo.getIndexes().length, 0);
+
+    assert.commandWorked(db1.copyDatabase('copydatabase_no_id_index', 'copydatabase_no_id_index2'));
+
+    assert.eq(db1.foo.count(), 1);
+    assert.eq(db1.foo.getIndexes().length, 0);
+    assert.eq(db2.foo.count(), 1);
+    assert.eq(db2.foo.getIndexes().length, 0);
+})();
\ No newline at end of file
diff --git a/src/mongo/db/cloner.cpp b/src/mongo/db/cloner.cpp
index 91fa3d94b75..9d4a5165424 100644
--- a/src/mongo/db/cloner.cpp
+++ b/src/mongo/db/cloner.cpp
@@ -618,7 +618,15 @@ Status Cloner::copyDb(OperationContext* txn,
             uassert(18645, str::stream() << "database " << toDBName << " dropped during clone", db);
 
             Collection* c = db->getCollection(to_name);
-            if (c && !c->getIndexCatalog()->haveIdIndex(txn)) {
+            bool autoIndexId = true;
+            auto status =
+                bsonExtractBooleanFieldWithDefault(options, "autoIndexId", true, &autoIndexId);
+            if (!status.isOK()) {
+                return status;
+            }
+
+            if (c && !c->getIndexCatalog()->haveIdIndex(txn) && autoIndexId) {
+
                 // We need to drop objects with duplicate _ids because we didn't do a true
                 // snapshot and this is before applying oplog operations that occur during the
                 // initial sync.