We have written web services/ callbacks in Laravel/Lumen for Azure graph APIs. Our web services are hosted in AWS ASG with Apache and PHP-FPM on the server. Minimum pool in ASG is 4 and max is set at 10. For database we are using MongoDB that also hosted in AWS parallel to Lumen APIs. We are using jenssegers/mongodb package in Lumen for database connection/queries. We are receiving huge number of requests from Azure per second. Our project is related to email/mail boxes so traffic/requests are coming continuously upto ~60K/sec.
The problem is our MongoDB connections are increasing rapidly reaching upto 40K. How to manage these connections? As MongoDB connection limit is 50K and we are also expecting increase in traffic from Azure in future with product expansion.
Earlier PHP-FPM was setup with dynamic PM, we have now changed it to ondemand PM with following configs
pm = ondemand
pm.max_children = 100
pm.process_idle_timeout = 20s;
pm.min_spare_servers = 10
pm.max_spare_servers = 20
pm.start_servers = 10;
This has decrease connection count but not significantly (around ~5K).