From bc85a804edfb36fab88ac6088ee92a9613dced45 Mon Sep 17 00:00:00 2001
From: Jordi Serra Torrens <jordi.serra-torrens@mongodb.com>
Date: Mon, 26 Feb 2024 13:15:53 +0000
Subject: [PATCH] Repro resharding and txn

---
 jstests/sharding/repro.js                     | 60 +++++++++++++++++++
 .../resharding_coordinator_service.cpp        |  7 +++
 2 files changed, 67 insertions(+)
 create mode 100644 jstests/sharding/repro.js

diff --git a/jstests/sharding/repro.js b/jstests/sharding/repro.js
new file mode 100644
index 00000000000..a9bdac31879
--- /dev/null
+++ b/jstests/sharding/repro.js
@@ -0,0 +1,60 @@
+import {configureFailPoint} from "jstests/libs/fail_point_util.js";
+
+(function() {
+'use strict';
+const dbName = 'test';
+const collName = 'foo';
+const ns = dbName + '.' + collName;
+
+const st = new ShardingTest({mongos: 1, shards: 2});
+const db = st.s.getDB(dbName);
+const coll = db[collName];
+
+assert.commandWorked(st.s.adminCommand({shardCollection: ns, key: {x: 1}}))
+assert.commandWorked(st.splitAt(ns, {x: 0}));
+assert.commandWorked(st.moveChunk(ns, {x: -1}, st.shard0.shardName));
+assert.commandWorked(st.moveChunk(ns, {x: 1}, st.shard1.shardName));
+
+assert.commandWorked(coll.insertMany([{x: -1, y: 0}, {x: 1, y: 0}]));
+
+// Set fpto block resharding after commit on configsvr but before commit on shards.
+const fp = configureFailPoint(st.configRS.getPrimary(),
+                              "reshardingPauseBeforeTellingParticipantsToCommit");
+
+// On parallel shell, start resharding
+const joinResharding = startParallelShell(() => {
+    assert.commandWorked(db.adminCommand({reshardCollection: 'test.foo', key: {y: 1}}));
+}, st.s.port);
+
+// Await configsvr to have done its part of the commit.
+fp.wait();
+
+const session = st.s.startSession();
+const sessionDB = session.getDatabase(dbName);
+const sessionColl = sessionDB.getCollection(collName);
+const sessionColl2 = sessionDB.getCollection('otherColl');
+
+// Make sure the session knows of a clusterTime inclusive of the resharding operation up to the
+// commit on the confisvr.
+assert.commandWorked(sessionColl2.insert({a: 1}));
+
+// Start txn.
+session.startTransaction({readConcern: {level: 'snapshot'}});
+assert.eq(1, sessionColl2.find().itcount());
+
+// Unset fp and wait for resharding to finish.
+fp.off();
+joinResharding();
+
+// Make sure the router is aware of the new (post-resharding) routing table for test.foo
+assert.eq(2, coll.find({y: 0}).itcount());
+
+// Query / write to coll. Check results.
+const docs = sessionColl.find({y: 0}).toArray();
+session.commitTransaction();
+
+jsTest.log("--DEBUG-- docs: " + tojson(docs));
+assert.eq(2, docs.length);  // BUG: Fails here! The query only sees one doc.
+
+st.stop();
+})();
diff --git a/src/mongo/db/s/resharding/resharding_coordinator_service.cpp b/src/mongo/db/s/resharding/resharding_coordinator_service.cpp
index 2aa1954cc52..ab05f6082f6 100644
--- a/src/mongo/db/s/resharding/resharding_coordinator_service.cpp
+++ b/src/mongo/db/s/resharding/resharding_coordinator_service.cpp
@@ -140,6 +140,7 @@ MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeInitializing);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeCloning);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeBlockingWrites);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeDecisionPersisted);
+MONGO_FAIL_POINT_DEFINE(reshardingPauseBeforeTellingParticipantsToCommit);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeRemovingStateDoc);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeCompletion);
 MONGO_FAIL_POINT_DEFINE(reshardingPauseCoordinatorBeforeStartingErrorFlow);
@@ -2760,6 +2761,12 @@ void ReshardingCoordinator::_tellAllDonorsToRefresh(
 
 void ReshardingCoordinator::_tellAllParticipantsToCommit(
     const NamespaceString& nss, const std::shared_ptr<executor::ScopedTaskExecutor>& executor) {
+    {
+        auto opCtx = _cancelableOpCtxFactory->makeOperationContext(&cc());
+        reshardingPauseBeforeTellingParticipantsToCommit.pauseWhileSetAndNotCanceled(
+            opCtx.get(), _ctHolder->getAbortToken());
+    }
+
     auto opts = createShardsvrCommitReshardCollectionOptions(
         nss, _coordinatorDoc.getReshardingUUID(), **executor, _ctHolder->getStepdownToken(), {});
     opts->cmd.setDbName(DatabaseName::kAdmin);
-- 
2.34.1