Question

I’d like some help understanding the use of pheanstalk (php beanstalk client). I have a PHP program that is executed on the server when form data is sent to it. The PHP program should then package the form data as a JSON structure and send it to a backend server process.

The thing I don’t understand is the connection to the beanstalkd server. Should I create a new Pheanstalk() object each time the PHP program executes – in which case, am I incurring the cost of creating the connection. When is the connection closed (since there is no close() method in pheanstalk)?

If the connection is persistent, is it shared among all executions of the PHP program, in which case, what happens in the case of concurrent hits? Thanks for any help.

Answer 1

Yes, you will have to create a new connection with Pheanstalk (or any other library) each time you start the program, since PHP starts each one fresh. The overhead is tiny though.

The Beanstalkd process is optimised to easily handle a number of connections, and will act on them atomically – you won’t get a duplicate job, unless you put two of the same in there (and even then, they would have different job-ID’s).

Pheanstalk doesn’t even send data to the daemon any information (including opening the connection) until the first command is sent. It’s for this reason that you can’t tell if the daemon is even alive till you actively make a request (in my tests, I get the list of current tubes). If you kept re-using the instantiated class in the running program, then it would keep reusing it of course.

There’s no formal close(), but unset($pheanstalk) would do the same thing, running the destructor. Again, the call is program so transient and the daemon can keep so many concurrent connections open if it’s allowed to, that it’s not an issue – and it will be shut down as the program itself does.

In short, don’t worry. The overhead of connecting and sending data into, or out of, Beanstalkd will probably be a tiny fraction of any work that is done by the worker, or producer, in generating the request/response.