the Server Resets the Connection when the Client-cpp Inserts Data into a Narrow Table

**Project**
Our project involves using ClickHouse to efficiently store and query large amounts of data. We want to do this using more flexible narrow tables.

**Describe the situation**
I encountered a problem while attempting to insert data into a narrow table using the ClickHouse-CPP client. When the number of rows is large (e.g., 1,000,000 rows), I receive a "Connection reset by peer" error. The error manifests with 1,000,000 rows but not with 500,000 rows.
Known information:
1. The client successfully transmitted a certain amount of data, after which the server reset the connection
2. The data was correctly written to the Block buffer
3. No other errors other than system errors were caught (such as the ProtocolError thrown by the client when the outputStream cannot be successfully written)
It can basically be ruled out that it is a client error and I suspect this issue may be related to network transmission or server-side configuration when inserting a large amount of data.

**How to reproduce**
* **ClickHouse server version:** 23.9.2.56
* **Interface:** ClickHouse-CPP client
* **Non-default settings:** None
* **`CREATE TABLE` statement:**

```sql
CREATE TABLE IF NOT EXISTS database.table_name
(date Date32, fname FixedString(18), symbol FixedString(8), time DateTime64, value Int32)
ENGINE = ReplacingMergeTree()
ORDER BY (date, fname, symbol, time);
```

* **Sample demo:**

```cpp
#include
#include
#include
#include
#include
#include

int main() {
const std::string _host = "host";
const std::string _user = "user";
const std::string _passwd = "password";
const std::string _database = "database";
const std::string _table = "table_name";
const std::string table_name = _database + "." + _table;

const int push_step = 1e6; // Number of rows per transfer
const int ColumnNumber = 99;

Client client(ClientOptions().SetHost(_host).SetUser(_user).SetPassword(_passwd).SetCompressionMethod(CompressionMethod::LZ4));

std::stringstream ss;
ss << "CREATE TABLE IF NOT EXISTS " << table_name
<< " (date Date32, fname FixedString(18), symbol FixedString(8), time DateTime64, value Int32"
<< ") ENGINE = ReplacingMergeTree() order by (date, fname, symbol, time)";
client.Execute(ss.str());

/******************** Prepare Data ********************/

std::random_device rd;
std::mt19937 gen(rd());
std::uniform_int_distribution dis(1, 10000);

std::shared_ptr date = std::make_shared();
std::shared_ptr symbol = std::make_shared(8);
std::shared_ptr fname = std::make_shared(18);
std::shared_ptr time = std::make_shared(3);
std::shared_ptr values = std::make_shared();

date->Reserve(push_step);
fname->Reserve(push_step);
symbol->Reserve(push_step);
time->Reserve(push_step);
values->Reserve(push_step);

for (auto i = 0; i < push_step; ++i) {
time_t cur_time;
std::time(&cur_time);
date->Append(cur_time);
symbol->Append("sh" + std::to_string(i).substr(0, 6));
time->Append(dis(gen));
}
for (auto i = 0; i < ColumnNumber; ++i) {
for (auto j = 0; j < push_step; ++j) {
fname->Append("f" + std::to_string(j));
values->Append(dis(gen));
}
}

/****************** Transfer Data **********************/

std::cout << "driver release start" << "\n";

Block block;
block.AppendColumn("date", date);
block.AppendColumn("symbol", symbol);
block.AppendColumn("fname", fname);
block.AppendColumn("time", time);
block.AppendColumn("value", values);
client.Insert(table_name, block);

std::cout << "driver release done" << "\n";

return 0;
}
```

**Additional context**
Upon execution, the program outputs "driver release start" and then terminates with the following error message:

```
terminate called after throwing an instance of 'std::system_error'
what(): fail to send 16 bytes of data: Connection reset by peer
Aborted
```

Could you advise on how to adjust ClickHouse or the client code to resolve this "Connection reset by peer" error in such a scenario? Thank you for your assistance.

まず、提供された C++ 再現プログラムを 500,000 行および 1,000,000 行で実行し、クライアントの出力を ClickHouse サーバーとネットワークのログと比較します。プロジェクトファイルやテストは指定されていないため、まず insert と transport のエントリポイントを見つけてから、リセットの原因がサーバー設定にあるのかクライアントの動作にあるのかを判断します。再現可能な原因を特定し、検証済みの調整を文書化できれば完了です。

索引モデルが issue の本文から書いたものです。