Method _checkpid

redis/connection.py:3076–3121

(self)

            pass

    def _checkpid(self) -> None:
        # _checkpid() attempts to keep ConnectionPool fork-safe on modern
        # systems. this is called by all ConnectionPool methods that
        # manipulate the pool's state such as get_connection() and release().
        #
        # _checkpid() determines whether the process has forked by comparing
        # the current process id to the process id saved on the ConnectionPool
        # instance. if these values are the same, _checkpid() simply returns.
        #
        # when the process ids differ, _checkpid() assumes that the process
        # has forked and that we're now running in the child process. the child
        # process cannot use the parent's file descriptors (e.g., sockets).
        # therefore, when _checkpid() sees the process id change, it calls
        # reset() in order to reinitialize the child's ConnectionPool. this
        # will cause the child to make all new connection objects.
        #
        # _checkpid() is protected by self._fork_lock to ensure that multiple
        # threads in the child process do not call reset() multiple times.
        #
        # there is an extremely small chance this could fail in the following
        # scenario:
        #   1. process A calls _checkpid() for the first time and acquires
        #      self._fork_lock.
        #   2. while holding self._fork_lock, process A forks (the fork()
        #      could happen in a different thread owned by process A)
        #   3. process B (the forked child process) inherits the
        #      ConnectionPool's state from the parent. that state includes
        #      a locked _fork_lock. process B will not be notified when
        #      process A releases the _fork_lock and will thus never be
        #      able to acquire the _fork_lock.
        #
        # to mitigate this possible deadlock, _checkpid() will only wait 5
        # seconds to acquire _fork_lock. if _fork_lock cannot be acquired in
        # that time it is assumed that the child is deadlocked and a
        # redis.ChildDeadlockedError error is raised.
        if self.pid != os.getpid():
            acquired = self._fork_lock.acquire(timeout=5)
            if not acquired:
                raise ChildDeadlockedError
            # reset() the instance for the new process if another thread
            # hasn't already done so
            try:
                if self.pid != os.getpid():
                    self.reset()
            finally:
                self._fork_lock.release()

    @deprecated_args(
        args_to_warn=["*"],