Recent issues with PHP-FPM

[-AJ-]

Our site recently started having issues with PHP-FPM.  The forums become unreachable, but logging into the server doesn't show any obvious services pegged at high CPU.  Restarting PHP-FPM or Apache seems to fix the problem.  Turning off PHP-FPM on the site seems to resolve the issue as well, but we obviously want to keep using PHP-FPM for performance if we can.  Additionally, the issue appeared before we switched from 4.2 to 4.3, so it doesn't appear related to the upgrade or switching to elasticsearch.

The site is ran on a server with CentOS 7/Cpanel/WHM and is currently using the default pool options for PHP-FPM which are Max Children 5, Process Idle Timeout 10, and Max Requests 20.  I don't want to blindly change these settings and possibly create a different issue.  The system has a 4-core Xeon with hyperthreading so 8 threads and 32GB of RAM that is not being overutilized.  The forums regularly have around 300-400 users/guests on them.

It had been stable for a long period of time before this started.

Can anyone advise on appropriate settings or next steps?  Should we just abandon PHP-FPM?

[Summit360]

Do you have a slew of bad bots? ahref etc?

If it helps I'm using php-fpm on a 8core/24gb vps with higher numbers online, so I'd stick with it. I don't use apache, so all my monitoring/tuning 'tricks' are nginx related.

Are you pressuring memory at all? 

So there are few changes, the upgrade, elastic search and some unknowns on traffic. Given it was happening before 4.3 I'd lay some cash on traffic, just pushing your processes over the edge, queuing etc.

[RevengeFNF]

5 hours ago, -AJ- said:

Our site recently started having issues with PHP-FPM.  The forums become unreachable, but logging into the server doesn't show any obvious services pegged at high CPU.  Restarting PHP-FPM or Apache seems to fix the problem.  Turning off PHP-FPM on the site seems to resolve the issue as well, but we obviously want to keep using PHP-FPM for performance if we can.  Additionally, the issue appeared before we switched from 4.2 to 4.3, so it doesn't appear related to the upgrade or switching to elasticsearch.

The site is ran on a server with CentOS 7/Cpanel/WHM and is currently using the default pool options for PHP-FPM which are Max Children 5, Process Idle Timeout 10, and Max Requests 20.  I don't want to blindly change these settings and possibly create a different issue.  The system has a 4-core Xeon with hyperthreading so 8 threads and 32GB of RAM that is not being overutilized.  The forums regularly have around 300-400 users/guests on them.

It had been stable for a long period of time before this started.

Can anyone advise on appropriate settings or next steps?  Should we just abandon PHP-FPM?

Its possible for you to change to Nginx + php-fpm?

[-AJ-]

No pressuring of memory at all.  It could be bad bots as we haven't ruled that out.  As far as converting to Nginx, we could do so but haven't.