For 1h my instance is hammered at 100% CPU. I just upscaled, but that didn’t change anything. I now have a hetzner server with 4 cores and 4 processes for each of the messenger commands. I figured they just have to process a long queue so that makes sense, but only if the queue has an end…
The only thing I found to debug what the messengers are doing is this: php bin/console messenger:stats
which yields the same result each time I run it:
----------- -------
Transport Count
----------- -------
async 100
async_ap 100
failed 2152
----------- -------
! [NOTE] Unable to get message count for the following transports: "sync".
Is there a better way to debug kbin messengers?
Ok, but messenger processes still operate in your PHP-fpm limits. The message queue only processes within those limits too, so you need to look at your fpm processes, see what the CPU / memory usage per process is like and then set the max accordingly.
You’re better off using htop to look at the actual processes running.
PHP-fpm had a memory per process setting and a few settings to control how many new processes can be spawned. You need to tweak this so that the CPU doesn’t hot 100%, I normally aim for a max of 90% but an average of around 50.