apache / apache/arrow-java

[JAVA] [ARROW-JDBC] Ssl Handler Exception - Swallowing a harmless 'connection reset by peer / broken pipe'

オープン
#117 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る
主要言語
Java
スター
94
フォーク
152
平均マージ
3日 16時間
マージ済み PR(30日)
11

説明

### Describe the usage question you have. Please include as many useful details as possible.

I have implemented flight sql server in java with tls protocol. Below is my server build code. The server internally connects with trino to fetch iceberg data and send it to the client.

```
Location location = Location.forGrpcTls("localhost", 32222);
final ArrowFlightServer example = new ArrowFlightServer(location);
Location listenLocation = Location.forGrpcTls("0.0.0.0", 32222);
File certificate = new File(SSL_CERT_PATH + "/cert.crt");
File priveteKey = new File(SSL_CERT_PATH + "/cert.key");
try (final BufferAllocator allocator = example.rootAllocator; final FlightServer server
= FlightServer.builder(allocator, listenLocation, example)
.useTls(certificate, priveteKey)
.headerAuthenticator(example.callHeaderInfo)
.build()) {
server.start();
server.awaitTermination();
}
```
I am using _adbc_driver_flightsql 0.9.0_ to connect to server from python client. Using below code, I am able to fetch data. I am using custom call headers in this scenario.
```
with flight_sql.connect(uri="grpc+tls://192.168.140.77:32222",
db_kwargs={
"adbc.flight.sql.rpc.call_header.username": "username",
"adbc.flight.sql.rpc.call_header.password": "Password",
"adbc.flight.sql.client_option.tls_skip_verify": "true"
}
) as conn:

with conn.cursor() as cur:
cur.execute(
"SELECT * FROM postgresql.public.tesla limit 10"
)

x = cur.fetch_arrow_table()
num_rows = x.num_rows
print(x)
```

But I get one server side exception when I change log level to DEBUG. Even though I get data as expected but in server logs I get below exception. When log level is set to INFO, I don't get this exception. Additionally, If I disable TLS and use non-secure method, I don't get this server side exception with log level set to DEBUG.

`2024-03-19 15:42:32 DEBUG io.netty.handler.ssl.SslHandler - [id: 0xa6229d8a, L:/192.168.140.77:32222 - R:/192.168.140.77:65174] Swallowing a harmless 'connection reset by peer / broken pipe' error that occurred while writing close_notify in response to the peer's close_notify`

Secondly if I change my python code to use standard jdbc username/password parameters like below,
```
with flight_sql.connect(uri="grpc+tls://192.168.140.77:32222",
db_kwargs={
"username": "username",
"password": "Password",
"adbc.flight.sql.client_option.tls_skip_verify": "true"
}
) as conn:
```
My python client code throws an exception and it does get the data at all.
```
Traceback (most recent call last):
File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/multiprocessing/process.py", line 314, in _bootstrap
self.run()
File "/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/multiprocessing/process.py", line 108, in run
self._target(*self._args, **self._kwargs)
File "/Users/a.bashir/Desktop/python_codes/newproj/venv/arrow_py_client.py", line 19, in execute_arrow_query
with flight_sql.connect(uri="grpc+tls://192.168.140.77:32222",
File "/Users/a.bashir/Desktop/python_codes/newproj/venv/lib/python3.10/site-packages/adbc_driver_flightsql/dbapi.py", line 120, in connect
conn = adbc_driver_manager.AdbcConnection(db, **(conn_kwargs or {}))
File "adbc_driver_manager/_lib.pyx", line 644, in adbc_driver_manager._lib.AdbcConnection.__init__
File "adbc_driver_manager/_lib.pyx", line 385, in adbc_driver_manager._lib.check_error
adbc_driver_manager._lib.OperationalError: ADBC_STATUS_UNKNOWN (1): [FlightSQL] [FlightSQL] flight: no authorization header on the response (Unknown; AuthenticateBasicToken)
```
Wonder what I am doing wrong.

### Component(s)

Java, Python

コントリビューションガイド

コントリビューションガイドを開く

調査の方向性

Java FlightServer の TLS セットアップと Python の flight_sql.connect の例から始め、その後 Netty SslHandler の close_notify ログと Flight SQL の認証動作を調査します。カスタムの call ヘッダーと標準の username/password パラメーターを比較します。期待される認証動作が文書化または修正され、無害な TLS シャットダウンログが適切に処理されれば完了です。

索引モデルが issue の本文から書いたものです。

評価

技術スタック
grpc, java, python
領域
authentication, backend-api-design, networking, security
issue の種類
バグ
難易度
4/5
見積もり時間
3〜5日
活発さ
停滞
明瞭さ
説明が足りない
初心者へのやさしさ
25/100

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。