Droplet with nginx + php-fpm memory have unusual peaks of memory usage
I have a droplet running a PHP application with nginx + php5-fpm + postgres, and also a piwik and a ganglia installations (php + mysql).
My application is running well in most part of the time, but I have unusual peaks of memory usage that I couldn't find any related issue. It happens just in exact hours when my server is receiving a bit more users than usual (like 40~60 online-users). But it is not a simple heavy load because of the number of online users. For example: the server was ok with more than 50 users online between 1:35 PM to 1:59 PM, and exactly at 14:00 PM the server had a memory peak and the nginx presented "502 Bad Gateway". One minute later I restarted both nginx and php-fpm, and again the server was running well with more than 50 users online.. until 15:00 PM, when the problem happens again.
I have checked all cron jobs and I did not find any process running.
It seems like php-fpm or nginx execute some task at these times, and it forces php-fpm to be unavailable for new requests (resulting a 502 error), does someone know what may be happening with my droplet?