I'd like to report to other users, that we are seeing a "404 not found" error for execution of PHP scripts on an average of 1 out of (approximately, on average) 70000 requests to the server across all of our CentOS 7.4 servers. This is happening at approximately the same tiny number of requests across different systems since the CentOS 7.4 upgrade was done.
It never occured before, however before the upgrade, once in a while a PHP 503 was delivered, which was caused by an exhausted number of PHP-FPM instances or crashed PHP-FPM services. These crashes or resource excessions became much more seldom after the CentOS 7.4 upgrade, while now we are monitoring the 404 issue instead.
It is unclear why this happens, but it seems that somewhere in the combination of Nginx and CentOS 7.4 and PHP-FPM service, every 1 out of (approximately, on average) 70000 request is lost and answered by backend Apache with that "404 not found". Doing further PHP script requests to the server the server responds correctly, but then, after (on average) another n attempts once in a while responds with a "404 not found" again. For the time being, our watchdog scripts catch the issue, restart the PHP-FPM module that is involved, so it is not a big issue.
But if anyone else sees the same issue, we'd still be interested to know, why it is happening.
It never occured before, however before the upgrade, once in a while a PHP 503 was delivered, which was caused by an exhausted number of PHP-FPM instances or crashed PHP-FPM services. These crashes or resource excessions became much more seldom after the CentOS 7.4 upgrade, while now we are monitoring the 404 issue instead.
It is unclear why this happens, but it seems that somewhere in the combination of Nginx and CentOS 7.4 and PHP-FPM service, every 1 out of (approximately, on average) 70000 request is lost and answered by backend Apache with that "404 not found". Doing further PHP script requests to the server the server responds correctly, but then, after (on average) another n attempts once in a while responds with a "404 not found" again. For the time being, our watchdog scripts catch the issue, restart the PHP-FPM module that is involved, so it is not a big issue.
But if anyone else sees the same issue, we'd still be interested to know, why it is happening.