murano-agent dead-lock on wrong execution-plan

[chenyujie]

During loop on process execution plan, there is no try/catch to protect the code sequence. So if there is an exception raised from get_execution_plan(such as a wrong json format, and there is no try/catch either in get_execution_plan), the message would be processed for ever because next(msg_iterator) would never be executed, and the agent would dead-lock at the message.

    def _loop_func(self, msg_iterator):
        result, timestamp = self._queue.get_execution_plan_result()
        if result is not None:
            if self._send_result(result):
                self._queue.remove(timestamp)
            return

        plan = self._queue.get_execution_plan()
        if plan is not None:
            LOG.debug("Got an execution plan '{0}':".format(str(plan)))
            self._run(plan)
            return

        next(msg_iterator)

[Stan Lagun (slagun)]

Could you please provide more information on your use cases? In normal usage you cannot get json decoding exceptions for jsons that were produced by the very same application. And if the corrupted json was put there by someone else hanging until it fixed back by the user might be a reasonable behavior

[chenyujie (gzyjchen)]

The problem is rare. In our environment, we write wrong code in resource template, which produce not a valid yaml then the agent is dead after the wrong request.