Showing posts with label http. Show all posts
Showing posts with label http. Show all posts

NodeJS, tune or disable HTTP agent pooling

When using NodeJS's HTTP module to query you might end-up having performance problems if you do some heavy HTTP querying.

By default when executing queries using the HTTP module NodeJS will use connection pooling with the maximum number of connection sockets being 5, so if you hit the bottleneck your HTTP queries will lag as they will be queued.

To my knowledge you have 2 options to "counter" this behavior :

  1. Increase the number of http sockets
  2. Disable the HTTP agent per request

1. Increase the number of http sockets

    var http = require('http');

    // increase the maximum number of http connections in the pool
    http.globalAgent.maxSockets = 100;

    //regular http request options
    var options = {
        port: 1337,
        hostname: '127.0.0.1',
        method: 'GET',
        path: '/test'
  };

  // make http query
  var req = http.request(options, function(res) {
       console.log('STATUS: ' + res.statusCode);
       console.log('HEADERS: ' + JSON.stringify(res.headers));
       res.setEncoding('utf8');
       res.on('data', function (chunk) {
          console.log('BODY: ' + chunk);
      });
   });



2.Disable the HTTP agent per request


    var http = require('http');
  
      //regular http request options
      var options = {
         port: 1337,
         hostname: '127.0.0.1',
         method: 'GET',
         path: '/test',
         agent:false //disable the nodejs agent (no connection pooling)
      };

      // make http query
      var req = http.request(options, function(res) {
       console.log('STATUS: ' + res.statusCode);
       console.log('HEADERS: ' + JSON.stringify(res.headers));
       res.setEncoding('utf8');
       res.on('data', function (chunk) {
          console.log('BODY: ' + chunk);
      });
   });

Please be aware that changing this options might have side effects on your host system such as higher memory and bandwidth use so use it wisely

REST load testing with siege

Siege is a pretty nifty linux utility for HTTP/HTTPS load testing it can be downloaded here : siege homepage
Siege allows you to write a list of URLs in a file (one URL per line) the utility will then parse the file and execute the load tests according to your configuration file .siegerc or the command's arguments.

The only thing I do not like about siege is the fact that I would like to have stats per URL; when siege parses the url file it will write a log file with global statics for the siege session

Something like so :

2012-09-21 12:03:52,   1813,     121.43,          19,       0.49,       14.93,        0.16,        7.38,    1813,       0

which to my taste is not very clear

While executing the tests siege prints out statistics that are not found in the siege log file (which in my case are more useful) but if you have multiple URLs you don't know for which URL the printed stats are.

So I decided to handle the situation differently using awk and a bash script

Note that I'm no Linux expert so the script could probably be better written but here is how I broke it down :

I have 3 files

  1. The urls.txt file (a txt file containing the URLs I want to test :
  2. #URL1
    http://myURL1/param1/param2.json
    http://myURL1/param3/param2.json
    #URL2
    http://myURL2/param3/param2.json
    http://myURL2/param3/param1.json
    
    
  3. The AWK script file
  4. !/#/{
    
       print "\n"
       print "**********************************************************"
       print "Testing URL : "$1
       print "**********************************************************"
       system ("echo Testing URL : "$1 " >> " SIEGE_OUTPUT " 2>&1")
       system ("siege " $1 "-v -b -r"REPS" -c"CONC" --mark="$1 " --log="SIEGE_LOG_FILE ">> "SIEGE_OUTPUT " 2>&1 ")
       print "\n"
    
    }
    
    
  5. The bash script containing all the configuration and calling the awk script
  6. #! /bin/sh
    #LOG_FILE=./siege-log
    URL_FILE=./urls.txt # file containing the URLs to handle
    USER=$(whoami)
    NOW=$(date +"%d-%m-%Y_%H-%M-%S")
    PWD=$(pwd)
    SIEGE_LOG_FILE=$PWD"/log_"$NOW".log"  #siege log's file
    SIEGE_OUTPUT=$PWD"/output_"$NOW".log" #where siege's output will be redirected
    CONC=50 #number of concurrent users
    REPS=1  # number of repetitions
    echo "loading URLS from the file : "$URL_FILE
    echo "Writing siege log into the file :" $SIEGE_LOG_FILE
    
    SET -- $CONC
    
    awk -f siege-benchmark.awk -v CONC=$CONC REPS=$REPS SIEGE_LOG_FILE=$SIEGE_LOG_FILE SIEGE_OUTPUT=$SIEGE_OUTPUT $URL_FILE
    
    
    
    

Now according to how your siegerc file is configured output can be a bit different I have the verbose mod off and benchmark mode on. Below is an excerpt from my log file

******************************************
Testing URL : http://myURL1/param1/param2.json
******************************************
** SIEGE 2.72
** Preparing 50 concurrent users for battle.
The server is now under siege...


Transactions:                     50 hits
Availability:                 100.00 %
Elapsed time:                   0.08 secs
Data transferred:               0.00 MB
Response time:                  0.04 secs
Transaction rate:             625.00 trans/sec
Throughput:                     0.03 MB/sec
Concurrency:                   24.62
Successful transactions:          50
Failed transactions:               0
Longest transaction:            0.07
Shortest transaction:           0.01

OSX show used ports or listening applications with their PID

On OSX you can display applications listening on a given port using the lsof the commands described below will show listening application...