When I simulate 100 users to send http requests to my bot at the same time, I find that the requests are queued. Only when the bot finishes processing one request can it start processing the next one. This is very slow. How can I handle it? Thanks.
this is due to the
Sanic server which - if none of the processed requests is blocked - processes requests one after the other. Also, if you are sending multiple requests for the same conversation that’s desired behavior. What do you mean “very slow”? Usually you also don’t have a 100 conversations at the same second, but rather a bit spread out.
Hi Tobias, we have a little different, but related issue. We use RASA connected with FB Messenger and the story seems mixed when our users send multiple messages in a very short time period (within a few seconds).
Does RASA core handle concurrency in intent classification and dialog control? Do we need to configure something to deal with multiple users on FB Messenger? BTW, our application server runs on Heroku and Clear DB(RDBMS).