Welcome to the LimeSurvey Community Forum

Ask the community, share ideas, and connect with other LimeSurvey users!

Nginx HTTP 499 errors with LimeSurvey 5.6.56 under concurrent load

More
12 hours 39 minutes ago #275112 by nayan9718
Please help us help you and fill where relevant:
Your LimeSurvey version: 5.6.56
Own server or LimeSurvey hosting:
Survey theme/template:
==================Hi Everyone,I am facing an issue with Nginx HTTP 499 errors when running LimeSurvey under concurrent participant load.Environment
  • LimeSurvey: 5.6.56
  • PHP: PHP-FPM
  • PHP-FPM workers: 192
  • Web server: Nginx
  • Database: PostgreSQL
  • Architecture: Nginx is running as a proxy in front of LimeSurvey
Concurrent LoadWe are currently seeing approximately:
  • 40 concurrent participants when one service is running
  • 80 concurrent participants when two services are running in parallel
When the concurrent load increases, we start seeing HTTP 499 errors in the Nginx access logs.Nginx Configuration
Code:
proxy_read_timeout 720s;
proxy_connect_timeout 720s;
proxy_send_timeout 720s;
 
underscores_in_headers on;
 
proxy_buffer_size 32k;
proxy_buffers 8 16k;
IssueWe are trying to understand why Nginx is returning HTTP 499 under this level of concurrent load.As I understand it, HTTP 499 is an Nginx-specific status code indicating that the client closed the connection before Nginx could return a response.However, I would like to determine whether the 499 errors could be caused by:
  • Slow LimeSurvey/PHP-FPM processing
  • PHP-FPM worker exhaustion or configuration
  • PostgreSQL query/connection performance
  • LimeSurvey session handling
  • Nginx reverse-proxy configuration
  • Load balancer/client timeout
  • Network issues
  • High CPU or memory utilization
  • Too many PHP-FPM workers causing resource contention
PHP-FPMWe currently have 192 PHP-FPM workers configured.Could having such a high number of workers cause resource contention, particularly if individual LimeSurvey requests consume significant CPU or memory?What PHP-FPM settings would you recommend checking for this type of workload?PostgreSQLThe application is running on PostgreSQL.Are there any specific PostgreSQL metrics or settings that we should monitor to determine whether slow database queries or connection limits are contributing to the delayed responses?Any suggestions or guidance on how to troubleshoot these 499 errors under concurrent LimeSurvey load would be greatly appreciated.Thank you.

Please Log in to join the conversation.

Moderators: holchtpartner

Lime-years ahead

Online-surveys for every purse and purpose