AFAIK In Django 1.6 the cursor is shared between requests and open a new connection when needed. However I couldn't find any information about how connection.cursor() method works and how can I should use connection.cursor() method in different processes in documentation.
Most code I have seen doesn't close cursor object returned from connection.cursor() so I assume that Django automatically close the cursor object returned from connection.cursor() when the request is finished, but what if I call connection.cursor() in a different process? Should I close that cursor when terminating the process or the wrapper function of cursor automatically kills itself?
Thread safety depends on db connection implementation, see Python Database API Specification v2.0 for possible cases:
threadsafety:Integer constant stating the level of thread safety the interface supports. Possible values are:threadsafety Meaning 0 Threads may not share the module. 1 Threads may share the module, but not connections. 2 Threads may share the module and connections. 3 Threads may share the module, connections and cursors.
for example, MySQL, PostgreSQL and Oracle:
>>> import MySQLdb
>>> MySQLdb.threadsafety
1
>>> import psycopg2
>>> psycopg2.threadsafety
2
>>> import cx_Oracle
>>> cx_Oracle.threadsafety
2
So it seems that no django supported DB connection is thread safe on cursor level.
Normally, cursors are closed with __del__ method, i.e. when garbage collected, or when thread is normally terminated (as contrary to when it's killed, as daemon threads may end up). I recall some bugs in cursors handling on specific platforms, namely some issues with oracle's famous ORA-01000 Too many cursors error. Quick google search revealed a bug for Jython, but I am quite sure I had issues in CPython as well. See also this ticket