diff --git a/env/env_test.cc b/env/env_test.cc
index 98f73d9a9..4f31dbd14 100644
--- a/env/env_test.cc
+++ b/env/env_test.cc
@@ -1180,6 +1180,99 @@ TEST_P(EnvPosixTestWithParam, MultiRead) {
   }
 }
 
+TEST_F(EnvPosixTest, MultiReadNonAlignedLargeNum) {
+  // In this test we don't do aligned read, wo it doesn't work for
+  // direct I/O case.
+  EnvOptions soptions;
+  soptions.use_direct_reads = soptions.use_direct_writes = false;
+  std::string fname = test::PerThreadDBPath(env_, "testfile");
+
+  const size_t kTotalSize = 81920;
+  std::string expected_data;
+  Random rnd(301);
+  test::RandomString(&rnd, kTotalSize, &expected_data);
+
+  // Create file.
+  {
+    std::unique_ptr<WritableFile> wfile;
+    ASSERT_OK(env_->NewWritableFile(fname, &wfile, soptions));
+    ASSERT_OK(wfile->Append(expected_data));
+    ASSERT_OK(wfile->Close());
+  }
+
+  // More attempts to simulate more partial result sequences.
+  for (uint32_t attempt = 0; attempt < 25; attempt++) {
+    // Right now kIoUringDepth is hard coded as 256, so we need very large
+    // number of keys to cover the case of multiple rounds of submissions.
+    // Right now the test latency is still acceptable. If it ends up with
+    // too long, we can modify the io uring depth with SyncPoint here.
+    const int num_reads = rnd.Uniform(512) + 1;
+
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
+        "PosixRandomAccessFile::MultiRead:io_uring_result", [&](void* arg) {
+          if (attempt > 5) {
+            // Improve partial result rates in second half of the run to
+            // cover the case of repeated partial results.
+            int odd = (attempt < 15) ? num_reads / 2 : 4;
+            // No failure in first several attempts.
+            size_t& bytes_read = *static_cast<size_t*>(arg);
+            if (rnd.OneIn(odd)) {
+              bytes_read = 0;
+            } else if (rnd.OneIn(odd / 2)) {
+              bytes_read = static_cast<size_t>(
+                  rnd.Uniform(static_cast<int>(bytes_read)));
+            }
+          }
+        });
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
+
+    // Generate (offset, len) pairs
+    std::set<int> start_offsets;
+    for (int i = 0; i < num_reads; i++) {
+      int rnd_off;
+      // No repeat offsets.
+      while (start_offsets.find(rnd_off = rnd.Uniform(81920)) != start_offsets.end()) {}
+      start_offsets.insert(rnd_off);
+    }
+    std::vector<size_t> offsets;
+    std::vector<size_t> lens;
+    // std::set already sorted the offsets.
+    for (int so: start_offsets) {
+      offsets.push_back(so);
+    }
+    for (size_t i = 0; i < offsets.size() - 1; i++) {
+      lens.push_back(static_cast<size_t>(rnd.Uniform(static_cast<int>(offsets[i + 1] - offsets[i])) + 1));
+    }
+    lens.push_back(static_cast<size_t>(rnd.Uniform(static_cast<int>(kTotalSize - offsets.back())) + 1));
+    ASSERT_EQ(num_reads, lens.size());
+
+    // Create requests
+    std::vector<std::string> scratches;
+    scratches.reserve(num_reads);
+    std::vector<ReadRequest> reqs(num_reads);
+    for (size_t i = 0; i < reqs.size(); ++i) {
+      reqs[i].offset = offsets[i];
+      reqs[i].len = lens[i];
+      scratches.emplace_back(reqs[i].len, ' ');
+      reqs[i].scratch = const_cast<char*>(scratches.back().data());
+    }
+
+    // Query the data
+    std::unique_ptr<RandomAccessFile> file;
+    ASSERT_OK(env_->NewRandomAccessFile(fname, &file, soptions));
+    ASSERT_OK(file->MultiRead(reqs.data(), reqs.size()));
+
+    // Validate results
+    for (int i = 0; i < num_reads; ++i) {
+      ASSERT_OK(reqs[i].status);
+      ASSERT_EQ(Slice(expected_data.data() + offsets[i], lens[i]).ToString(true),
+                reqs[i].result.ToString(true));
+    }
+
+    ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
+  }
+}
+  
 // Only works in linux platforms
 #ifdef OS_WIN
 TEST_P(EnvPosixTestWithParam, DISABLED_InvalidateCache) {