Refresh pool can break any code at any time

As shown in #11400 (closed), the pool can be refresh at any time since rev 32cd371b3e9c and so code using pool instance can also in many cases:

• The pool instance has no more objects, this can be solved by storing the _pool on the instance
• The pool singleton for a database can be stopped (not initialized) at any time. I see at least two possible solutions:
  • Keep the previous pool instance and use it as fallback if the current one is not initialized
  • Raise a specific exception so the dispatcher, cron or worker can retry it with a initialized pool.

assigned to @ced

added trytond type::crash + 1 deleted label

Maybe a better solution would probably that we store in a thread local the singleton instance of the Pool at the beginning of a request. The Pool instance must become immutable and so when a Pool instance must be refreshed, it is a new Pool instance that is created and that will replace the existing Pool as global singleton. This way the other threads that have already been started, keep using the former pool until they end.

• Make Pool.init works on a copy of the Pool instance
  • Use a threadlocal to store the new pool instance initialized such that call to Pool() in __setup__ return the initialized pool.
• Pool.refresh should not call Pool.stop but store that it must be refreshed
  • Pool.database_list must remove the database to refresh
• Transaction.start must store on Pool in a threadlocal the current pool instance to use. So the pool singleton is always the same for the duration of the transaction. (only for the first nested transaction)
  • Transaction.stop must remove the pool instance. (only for the last nested transaction)

Pool.refresh should not call Pool.stop but store that it must be refreshed

Indeed once we have one instance of pool per thread with a copy of the actual pooled objects, we can stop the Pool as existing pool instance will keep their local copy of the objects.

Transaction.start must store on Pool in a threadlocal the current pool instance to use.

We do not need to use the transaction if we return a singleton of pool per thread and database.

We do not need to use the transaction if we return a singleton of pool per thread and database.

This is not completely true because we do not know when to refresh the _pool. But I do not think it is a problem if the pool is changed during a transaction as long as it is replaced by a complete new _pool.

marked #12871 (closed) as a duplicate of this issue

marked this issue as related to #12871 (closed)

mentioned in commit ba68db6818a9

mentioned in merge request !1150 (merged)

mentioned in commit 30429d744d29

closed with merge request !1150 (merged)

Apply 30429d744d29 , we've noticed that when try to logging in or refreshing the session, an error trace has been triggered:

Traceback (most recent call last):
  File "dd/wsgi.py", line 108, in dispatch_request
    self.check_request_size(request, max_request_size)
  File "dd/wsgi.py", line 88, in check_request_size
    if request.user_id:
       ^^^^^^^^^^^^^^^
  File "/functools.py", line 1001, in __get__
    val = self.func(instance)
          ^^^^^^^^^^^^^^^^^^^
  File "dd/protocols/wrappers.py", line 119, in user_id
    user_id = security.check(
              ^^^^^^^^^^^^^^^
  File "dd/security.py", line 83, in check
    Session = pool.get('ir.session')
              ^^^^^^^^^^^^^^^^^^^^^^
  File "dd/pool.py", line 165, in get
    return self._pool[type][name]
           ~~~~~~~~~~~~~~~~^^^^^^
KeyError: 'ir.session'

However, with the trytond-console, everything is working correctly.

Run trytond-cron with #11401 (closed)

1 140460226766528 [2024-03-19 09:22:57,095] INFO trytond.ir.cron cron started for "xxx"
Exception in thread Thread-55 (run):
Traceback (most recent call last):
  File "/usr/lib/python3.11/threading.py", line 1038, in _bootstrap_inner
    self.run()
  File "/usr/lib/python3.11/threading.py", line 975, in run
    self._target(*self._args, **self._kwargs)
  File "/tryton/trytond/trytond/ir/cron.py", line 164, in run
    with transaction.start(
         ^^^^^^^^^^^^^^^^^^
  File "/tryton/trytond/trytond/transaction.py", line 204, in start
    self.database.lock(self.connection, table)
  File "/tryton/trytond/trytond/backend/postgresql/database.py", line 457, in lock
    Identifier(table)))
    ^^^^^^^^^^^^^^^^^
  File "/venv/lib/python3.11/site-packages/psycopg2/sql.py", line 327, in __init__
    raise TypeError("SQL identifier parts must be strings")
TypeError: SQL identifier parts must be strings