0

I have a pretty good nginx configuration and I am trying to test the load for 1000 requests in 10 seconds for loading index.html page from nginx with https connection. For all the 1000 samples, latency and connect time is consistent. But the response time is ok for the first few samples, but its really really bad for the remaining samples. I am attaching the samples. Please see the first screenshot. You can see the entries are consistent. But in 2nd screenshot, you can find the response time getting larger and larger as it progresses. What could be the reason for this?????

  • is it because of the hardware port limit ( 32000 - 61000 ) ? But for virtual hosted machine, I am not allowed to edit this.

  • is there any file descriptor open limit for virtual hosted machine? You can see that I set ulimit -a and -n to be around 200000. Will this work for a virtual machine?

  • Do I have to make anymore adjustment in nginx configuration to make the response time consistent ?

  • I am trying this HTTPS connection, is it b/c of the fact that each connection taking time in encryption and system getting busy with cpu cycles ? I am just trying to understand whether the problem is in hardware / software level ?

My Hardware is Virtual Hosted. Below is the configuration

  • Architecture: x86_64
  • CPU op-mode(s): 32-bit, 64-bit
  • Byte Order: Little Endian
  • CPU(s): 16
  • Vendor ID: GenuineIntel
  • CPU family: 6
  • Model: 26
  • Stepping: 5
  • CPU MHz: 2266.802
  • BogoMIPS: 4533.60
  • Virtualization: VT-x

This is the nginx configuration:

server {
    listen      80;
    server_name xxxx;

    # Strict Transport Security
    add_header Strict-Transport-Security max-age=2592000;
    rewrite ^/.*$ https://$host$request_uri? permanent;
}

server {
    listen 443 ssl;
    server_name xxxx;

    location / {     
    try_files $uri $uri/ @backend;  
    }   

    ## default location ##
    location @backend {
    proxy_buffering off;
    proxy_pass    http://glassfish_servers;

    proxy_http_version 1.1;
    proxy_set_header Connection "";

    # force timeouts if the backend dies
    proxy_next_upstream error timeout invalid_header http_500 http_502 http_503 http_504;
    #proxy_redirect off;

    # set headers
    proxy_set_header    Host            $host;
    proxy_set_header    X-Real-IP       $remote_addr;
    proxy_set_header    X-Forwarded-For $proxy_add_x_forwarded_for;
    proxy_set_header    X-Forwarded-Proto https;
    }

    ssl_certificate /etc/nginx/ssl/ssl-bundle.crt;
    ssl_certificate_key /etc/nginx/ssl/xxxx.key;

    ssl_session_cache shared:SSL:20m;
    ssl_session_timeout 10m;

    ssl_prefer_server_ciphers       on;
    ssl_protocols                   TLSv1 TLSv1.1 TLSv1.2;
    ssl_ciphers                     ECDH+AESGCM:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:ECDH+3DES:DH+3DES:RSA+AESGCM:RSA+AES:RSA+3DES:!aNULL:!MD5:!DSS;                   

}

enter image description here

enter image description here

This is the aggregate report I collected:

enter image description here

Kenshin
  • 1,030
  • 2
  • 12
  • 41

1 Answers1

1

Maybe it is not nginx issue. Depending on how you configure JMeter the load may be much higher than you expect it to be.

I would recommend the following: as 1000 requests per 10 seconds is 100 requests/second,

  1. Add Thread Group with

    • Threads: ~300
    • Ramp-up: 30
    • Loop count: Forever
    • Scheduler -> Duration -> i.e. 240 seconds
  2. Add HTTP Request to hammer index.html
  3. Add Constant Throughput Timer to limit the load to 100 requests / second - it will be 6000 requests per minute

  4. Provide JMeter enough Java HEAP space - depending on your OS locate the following line in jmeter or jmeter.bat file:

    HEAP="-Xms512m -Xmx512m"

    and adjust values to start with 50% and end with 80% of your total physical RAM

  5. Run JMeter in non-GUI mode with all listeners disabled like:

    jmeter -n -t /path/to/your/testplan.jmx -l /path/to/testresults.jtl
  6. Once test finishes open JMeter GUI, add listener of your choice (Aggregate Report is quite good) and analyse results

You can also take a look into JMeter Plugins project, it provides i.e. PerfMon Metrics Collector so you will be able to monitor your server health during the load test and a number of extended graphs to visualise your test or Taurus which is a wrapper on JMeter and other load testing tools which simplifies configuration and execution processes and makes results analysis easier.

Dmitri T
  • 159,985
  • 5
  • 83
  • 133