-
Type:
Bug
-
Resolution: Unresolved
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
Replication - Oplog and Recovery
-
ALL
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Summary
A mongod 9.0.5 shard primary aborted with fatal assertion 12436504, "Attempted to reserve optime without a declared write intent", raised from LocalOplogInfo::getNextOpTimes. The node did not recover and stayed down until the test fixture tore the cluster down ~12 minutes later.
Found in MongoDB CI for the mongosync (Cluster-to-Cluster Sync) project, in a suite that repeatedly kills and restarts mongos processes against a 2-shard cluster while a migration is running.
Log
{"t":{"$date":"2026-11-10T20:55:36.731+00:00"},"s":"F","c":"REPL","id":12436504,"svc":"S","ctx":"conn133",
"msg":"Attempted to reserve optime without a declared write intent","attr":{"opCtx":1960282}}
{"t":{"$date":"2026-11-10T20:55:36.731+00:00"},"s":"F","c":"ASSERT","id":23089,"svc":"S","ctx":"conn133",
"msg":"Fatal assertion","attr":{"msgid":12436504,
"location":"src/mongo/db/repl/local_oplog_info.cpp:141:52:
std::vector<OplogSlot> mongo::LocalOplogInfo::getNextOpTimes(OperationContext *, std::size_t, std::size_t)"}}
A second thread (conn237) hit the same assertion 9 ms later.
Stack
From the BACKTRACE dump (upper frames only; the lower SBE frames in the dump appear mis-symbolized):
mongo::fassert_detail::failed(mongo::fassert_detail::MsgId, mongo::WrappedStdSourceLocation) mongo::repl::logOp(mongo::OperationContext*, mongo::repl::MutableOplogEntry*) mongo::OpObserver::onOpMessage(mongo::OperationContext*, mongo::BSONObj const&) mongo::TransactionParticipant::performNoopWrite(mongo::OperationContext*, std::basic_string_view<char, ...>)
So the no-op write issued by TransactionParticipant::performNoopWrite reaches logOp without a declared write intent, and the optime reservation fasserts.
Environment
- mongod 9.0.5-patch-6ae3d8619708df0007da8205, Amazon Linux 2023 arm64
- Sharded cluster, 2 shards; the crashing node was the shard0 primary (port 20001)
- Workload: mongosync Cluster-to-Cluster migration (9.0 -> 9.0) running jscore passthrough tests, with a background hook killing/restarting a random mongos roughly every 30s (TERMINATE and KILL alternating, ~17s down each time)
Impact
The primary aborts (exit code -6) and does not come back, so the shard has no primary for the rest of the run. Not known to be reachable outside a test environment, but it is an unconditional fassert on a normal no-op write path.
Possibly related
SERVER-139006 ("Don't register a write intent for an operation that has been killed") touches the same write-intent machinery; the connection has not been verified.
Artifacts
Seen once so far. Tracked on the mongosync side as REP-8228.
- is related to
-
SERVER-139006 Don't register a write intent for an operation that has been killed
-
- Closed
-