Running a publisher-only perf test with predeclared queue

[rabbitmq]

Running a publisher-only perf test with predeclared queue

[acogoluegnes]

@lukebakken Right now the creation and the binding are in the same Java method. This method isn't called if -p is specified. Reverting #25 would make the binding happen.

Nevertheless, a direct-named exchange is created and is bound to the queue. I'd rather use the default exchange as you suggested.

[michaelklishin]

--predeclared by definition assumes you set your topology up out of band, any way the user likes. We can make publishers declare the queue but that's a little odd.

[lukebakken]

I do think something has changed since the example given in the docs (throughput-test-14 / "pre-populate a queue") implies that using --predeclared and --queue should be all you need to do to use a pre-existing queue.

[michaelklishin]

Hm, yes, my reading matches yours.

[lukebakken]

I'll see if I can figure it out tomorrow unless @acogoluegnes beats me to it 😃

[acogoluegnes]

I tried with a57ed36 (just after the initial import from the Java Client) and the -y0 -p -u "throughput-test-14" -s 1000 -C 1000000 --id "test-14" -f persistent declares the queue an configure it, so it's a regression. More digging.

[acogoluegnes]

It was introduced between 1.2.0 and 1.3.0, here.

It was a fix for #25. We shouldn't declare the queue but at least bind it. The code in configureQueues looks correct though.

[acogoluegnes]

I don't know how to reproduce #25 actually: as soon as --predeclared is used, queues are not created (they're bound though). I'd vote to reverse #25.

[lukebakken]

@acogoluegnes here is the most confusing behavior using the current PerfTest code (master):

• Start up RMQ and create a durable queue named throughput-test-14
• Start up PerfTest with these arguments:

  -y0 -p -u "throughput-test-14" -s 1000 -C 1000000 --id "test-14" -f persistent
  

I don't expect the queue to be created since I'm using the -p argument, but I do expect messages to be published using the default exchange with a routing key of throughput-test-14 so that they end up in the throughput-test-14 queue.

However, I'm not seeing this happen - when I run that command, the messages aren't routed so I assume they are being published without a routing key. What's even more odd is that if I add the -k throughput-test-14 argument they still aren't routed 🤔

[acogoluegnes]

Found an argument combination to reproduce #25: using --predeclared without specifying any queue. The auto-generated queue is then created. This was originally introduced in #5 (support for multiple queues). Only reverting #25 isn't enough then.