-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
-
Sharding EMEA
-
Fully Compatible
-
ALL
-
Sharding EMEA 2023-10-02, Sharding EMEA 2023-10-16, Sharding EMEA 2023-10-30, CAR Team 2023-11-13, CAR Team 2023-11-27, CAR Team 2023-12-11, CAR Team 2023-12-25
-
2
This is because the logic in ShardKeyPattern::extractShardKeyFromDoc will fill in BSON null for any value in the shard key pattern that is not found in the document.
This can cause a problem if a document to insert is passed to this method that does not already have an _id, and _id is in the shard key pattern. The document will then be routed to a shard based on an _id of BSON null. But once the document arrives on the shard, the shard will insert it, and give it a new _id, since it lacks one. And once it has a new _id, its value for the shard key has changed, and it may not be on the correct shard given its shard key.
This happens to not be a problem right now for the insert command, because mongos always adds _id to incoming insert command requests. But we noticed this problem with the new bulkWrite command where we weren't generating _id on mongos.
We will update the bulkWrite command to follow the same logic as insert to ensure there is always an _id, but it might be good to add some safeguard(s) around this. For example, maybe this method should error if _id is in the shard key pattern but not in the document.