-
Type: Bug
-
Resolution: Fixed
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
-
Cluster Scalability
-
Fully Compatible
-
ALL
-
Cluster Scalability 2024-09-02, Cluster Scalability 2024-10-28
-
0
In the internal_transactions_sharded_from_mongod_kill_sessions test, we run internal transactions and sometimes kill the sessions the transactions belong to. However, this abort is a best effort thing, and if the abortTransaction command runs before the command with the startTransaction: true argument in one of the shards then we will start the transaction after the abort and it will remain there until its lifetime expires. Because this operation is holding an IX lock, it will block any DDL command which will, in turn, block all other write operations.
This should be primarily a testing issue because our tests increase the transaction lifetime whereas in production, the transaction expiring would happen sooner and unblock other operations.
- is duplicated by
-
SERVER-93491 ClusterServerParameterRefresher can leave an open transaction on the CSRS after mongos shutdown.
- Closed
- is related to
-
SERVER-91604 Kill any orphan txn on the participant in multi_collection_transaction_placement_conflict_workaround.js
- Closed
-
SERVER-94039 Consider making abortTransaction remember the abort if TransactionParticipant hasn't been created yet
- Open