Hyukjin Kwon created SPARK-61182:
------------------------------------
Summary: Connect over TLS when the connection string has a token
for a host other than localhost, take any UUID as the session ID, and default
the user ID to SPARK_USER or USER, as PySpark does
Key: SPARK-61182
URL: https://issues.apache.org/jira/browse/SPARK-61182
Project: Spark
Issue Type: Bug
Components: Connect
Affects Versions: connect-rust-4.3.0
Reporter: Hyukjin Kwon
Three connection-string parameters of the client differ from PySpark Connect's
ChannelBuilder:
1. A token implies TLS in PySpark (secure = use_ssl or token is not None):
sc://host/;token=T, or a token of SPARK_CONNECT_AUTHENTICATE_TOKEN, connects
over TLS with the token for any host other than localhost. The client refuses
it before any RPC unless use_ssl=true is set: "Refusing to send the
authentication token to '<host>' over an insecure connection".
2. PySpark checks a session_id with uuid.UUID(session_id, version=4), which
takes any well-formed UUID. The client takes only version-4 UUIDs, and raises
INVALID_SESSION_UUID_ID for a version-1 or nil UUID.
3. PySpark's user ID is that of the connection string, else of SPARK_USER, else
of USER, and it is sent in the user context of every request. The client sends
an empty user ID.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
