Is it possible that HTTPRL kicking off additional threads that call the webserver on additional, detatched threads could cause an overabundance of calls hitting apache and in turn cause it to hit MaxClient issues more frequently?

We're talking hypothetically of course but it seems like we'd be hitting apache more often in exchange for improving the load of a single active thread for the UX.

Comments

btopro created an issue.

mikeytown2’s picture

mikeytown2 commented

100% correct. In practice not as bad as you'd think because you're deferring something that would have ran in the current process to run in a new process. The only work that gets repeated is the drupal bootstrap. If you have a slow bootstrap then yes you might have an issue; we haven't had any apache issues though. If you're using something like httprl_call_user_func_array_cache() to cache a slow function and set the cache_lifetime_min to a value greater than zero then httprl will actually free up more apache workers since that php request is running faster and the expensive function that you cached is now not being ran.

btopro’s picture

btopro commented

Context: All authenticated traffic running on a multi-site. Seems like I've had to bump up my MaxClients a bunch in the last week since upgrading httprl / advagg from 24 to 40 to 80 (and jack up ram to handle). Not blaming this project, just looking for some leads as to what might be suddenly generating so many more apache workers