TCP connection not closing properly leading to dangling CLOSE_WAIT state and system overload

Riproduci entrambi i casi usando Python 3.10.15, databricks-sql-python 3.1.2, le chiamate sql.connect mostrate e netstat per osservare gli stati dei socket. L’attività è completata quando le connessioni si chiudono correttamente dopo il context manager o le chiamate esplicite a close, non rimane alcun socket CLOSE_WAIT trattenuto indefinitamente e le risorse del client vengono liberate.

Testcase 1

Python 3.10.15 (main, Sep  7 2024, 18:35:33) [GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from databricks import sql
>>> sql.__version__
'3.1.2'
>>> with sql.connect(server_hostname = "<HIDE>", http_path = "<HIDE>", access_token = "<HIDE>") as connection:
...     with connection.cursor() as cursor:
...             cursor.execute("select * from table")
...             column_names = [desc[0] for desc in cursor.description]
...             query_out = [dict(zip(column_names, row)) for row in cursor.fetchall()]
...             
... 
<databricks.sql.client.Cursor object at 0x7f88551c14e0>
>>> query_out
[{"sometime" : "something}]


On the sidelines, the network connection for this process (after with clause gets finished), remain in ESTABLISHED state and then moves to CLOSE_WAIT state indefinitely. 

Every 1.0s: netstat -apn | grep 1070099                                                                                          asinha-vm-ubuntu-20-04-1: Wed Nov 20 02:43:13 2024

tcp       25      0 10.5.220.4:58078        so:me:ip:add:443      CLOSE_WAIT  1070099/python3

Testcase 2

Python 3.10.15 (main, Sep  7 2024, 18:35:33) [GCC 9.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from databricks import sql
>>> sql.__version__
'3.1.2'
>>> connection = sql.connect(server_hostname = "<HIDE>", http_path = "<HIDE>", access_token = "<HIDE>")
>>> ## at this point a TCP connection got made as ESTABLISHED state ##
>>> cursor = connection.cursor()
>>> cursor.execute("select * from table")
<databricks.sql.client.Cursor object at 0x7f9d65db26b0>
>>> column_names = [desc[0] for desc in cursor.description]
>>> query_out = [dict(zip(column_names, row)) for row in cursor.fetchall()]
>>> cursor.close()
>>> connection.close()
>>> # even after close being called, the connection remains in ESTABLISHED state (for a long time) and then finally moves to CLOSE_WAIT state, which never changes.

Every 1.0s: netstat -apn | grep 1074808                                                                                          asinha-vm-ubuntu-20-04-1: Wed Nov 20 02:49:36 2024

tcp       25      0 10.5.220.4:53810        so:me:ip:add:443      CLOSE_WAIT  1074808/python3

Connections should be closed properly, and all client resources should be freed properly.