fasthttp has an internal default limit of 512 max connections per host. This causes network errors if you use a concurrency higher than 512. Made a change to set the fasthttp MaxConnsPerHost to be equal to the concurrency configured in the gobench execution.

Before:

$ gobench -c 750 -t 5 -u http://openresty/
Dispatching 750 clients
Waiting for results...
Requests:                          4842824 hits
Successful requests:                  5990 hits
Network failed:                    4836833 hits
Bad requests failed (!2xx):              0 hits
Successful requests rate:             1198 hits/sec
Read throughput:                   1435200 bytes/sec
Write throughput:                   105332 bytes/sec
Test time:                               5 sec

After:

$ ./gobench -c 750 -t 5 -u http://openresty/
Dispatching 750 clients
Waiting for results...
Requests:                            52296 hits
Successful requests:                 52296 hits
Network failed:                          0 hits
Bad requests failed (!2xx):              0 hits
Successful requests rate:            10459 hits/sec
Read throughput:                  11546956 bytes/sec
Write throughput:                   859345 bytes/sec
Test time:                               5 sec

I tried this gobench -u http://google.com/ -k=true -c 1 -r 1

which results in

Dispatching 1 clients
Waiting for results...

Requests:                                1 hits
Successful requests:                     0 hits
Network failed:                          0 hits
Bad requests failed (!2xx):              1 hits
Successful requests rate:                0 hits/sec
Read throughput:                       552 bytes/sec
Write throughput:                       82 bytes/sec
Test time:                               1 sec

Running gobench -u http://localhost:3000/ -k=true -c 1 -r 1 gives successful results as expected.

Thanks in advance.

In high throughput low elapsed time situation, the request rate accuracy is very low.

For example,

$ ./gobench -c=1000 -r=10 -tc=10000 -tr=10000 -tw=10000 -u="<SAMPLE URL>"
Dispatching 1000 clients
Waiting for results...

Requests:                            10000 hits
Successful requests:                 10000 hits
Network failed:                          0 hits
Bad requests failed (!2xx):              0 hits
Successfull requests rate:            5000 hits/sec
Read throughput:                  24880000 bytes/sec
Write throughput:                   720000 bytes/sec
Test time:                               2 sec

$ ./gobench -c=1000 -r=10 -tc=10000 -tr=10000 -tw=10000 -u="<SAMPLE URL>"
Dispatching 1000 clients
Waiting for results...

Requests:                            10000 hits
Successful requests:                 10000 hits
Network failed:                          0 hits
Bad requests failed (!2xx):              0 hits
Successfull requests rate:            3333 hits/sec
Read throughput:                  16586666 bytes/sec
Write throughput:                   480000 bytes/sec
Test time:                               3 sec

This patch shows RPS/throughput as float instead of integer.

It would be convinient to have ulimit set automatically by GoBench, instead of me running ulimit -n 200000 each time I open a terminal for running GoBench.

The following metrics appearing in Siege summary are missing in GoBench summary when the test is done:

• Concurrency
• Throughput in MB/sec
• Total bytes (they have a different name for that)

Three modifies:

1. support for the url expression. so we can generate different url in a single url. It's useful when we want to test a web service with different arguments each times,such as login id
2. statistic output changes a little
3. use a more memory efficient way to recv the http body

./gobench5 -c=25000 -d="postbig.txt" -k=true -r=1 -u="http://myurl" -tc=10000 -tw=10000 -tr=10000 -ta=300000

postbig = ~0.5mb Running this is parallel from two computers, resulted in one client terminated after 226 seconds, while the other is stuck. There are ~25k connection on the server side.

test@debian:~$ curl -6 "http://[::1]/"
<!DOCTYPE html>
<html>
<head>
<title>Welcome to nginx!</title>
<style>
    body {
        width: 35em;
        margin: 0 auto;
        font-family: Tahoma, Verdana, Arial, sans-serif;
    }
</style>
</head>
<body>
<h1>Welcome to nginx!</h1>
<p>If you see this page, the nginx web server is successfully installed and
working. Further configuration is required.</p>

<p>For online documentation and support please refer to
<a href="http://nginx.org/">nginx.org</a>.<br/>
Commercial support is available at
<a href="http://nginx.com/">nginx.com</a>.</p>

<p><em>Thank you for using nginx.</em></p>
</body>
</html>

test@debian:~$ ./gobench -t 5 -u "http://[::1]/"
Dispatching 100 clients
Waiting for results...

Requests:                          1771986 hits
Successful requests:                     0 hits
Network failed:                    1771985 hits
Bad requests failed (!2xx):              0 hits
Successful requests rate:                0 hits/sec
Read throughput:                         0 bytes/sec
Write throughput:                        0 bytes/sec
Test time:                               5 sec
test@debian:~$ 

Hi. Sorry, I'm new to github, not sure if I did this by the book. I found a major rounding error in your calculations of requests/second, because you were dividing by the rounded number of seconds instead of the actual time passed. That would give you very incorrect results for short tests. I fixed it and I pushed it here:

https://github.com/i-fail/gobench-windows

I also needed a windows binary, which your repo didn't have, added that too.

Thank you for this useful tool!

Hello,

Please i have an application composed from two functions one is a get and the other one add function with post j son data , in this case how can i call this application using gobench.

Thank you.

I just remove the dependency on: err := proc.Signal(os.Interrupt) which is not supported in windows. This should resolve issue: https://github.com/cmpxchg16/gobench/issues/19