When I wrote Boom a few years ago, to replace Apache Bench (ab) in my work, Python was still maturing its async mechanisms, and gevent was a popular way to build an asynchronous HTTP client.
Fast forward 2020, asynchronous programming is now quite simple in Python and can be done natively, using async and await directives. I wrote Molotov, a load testing framework based on Python's asyncio and on the aiohttp client, and always wanted to rewrite Boom using Molotov, since it's just a very simple Molotov load test scenario: hitting a single HTTP(s) endpoint.
Instead of refactoring Boom, which is Python 2 compatible, I started a new tool called Salvo, which is Python 3 only. Salvo takes back Boom's command line interface and runs a Molotov script under the hood, and prints results the same way Boom is doing.
It's using a single process and runs as many concurrent "workers" as you want, which is enough to smoke test one endpoint. The philosophy is that anything more complex should be done using a full Molotov script.
To use Salvo, install it using pip and then send some load to an endpoint. Running with no option will perform a single hit
$ pip install salvo $ salvo https://example.com -------- Server info -------- Server Software: ECS (bsa/EB12) -------- Running 1 queries - concurrency 1 -------- [=================================================================>] 100% -------- Results -------- Successful calls 1 Total time 0.4430 s Average 0.4430 s Fastest 0.4430 s Slowest 0.4430 s Amplitude 0.0000 s Standard deviation 0.000000 Requests Per Second 2.26 Requests Per Minute 135.44 -------- Status codes -------- Code 200 1 times. Want to build a more powerful load test ? Try Molotov ! Bye!
You can then try to run more load. In the example below, we run 100 requests, across 10 concurrent workers:
$ salvo https://example.com -n 10 -c 10 -------- Server info -------- Server Software: ECS (bsa/EB12) -------- Running 10 queries - concurrency 10 -------- [=================================================================>] 100% -------- Results -------- Successful calls 100 Total time 14.7801 s Average 0.1478 s Fastest 0.1049 s Slowest 0.5314 s Amplitude 0.4265 s Standard deviation 0.115398 Requests Per Second 6.77 Requests Per Minute 405.95 -------- Status codes -------- Code 200 100 times. Want to build a more powerful load test ? Try Molotov ! Bye!
You can also use --duration to run the test for a given number of seconds, and use --json if you want to get the results in a JSON output.
In the example below, we also use --quiet to get only the JSON output, of a 10 seconds run using 10 workers:
$ salvo https://example.com -d 10 -c 10 --json --quiet
{"count": 774, "total_time": 102.28760695457458, "rps": 7.566899090167683,
"avg": 0.13215453094906277, "min": 0.10414314270019531,
"max": 0.8381381034851074, "amp": 0.7339949607849121,
"stdev": 0.09841954760413639,
"rpm": 454.013945410061, "server": {"software": "ECS (bsa/EB18)"}}
Salvo has a few basic customization features (see --help) but I won't extend them, the plan is to keep it as simple as possible and invite developers that need more features to use Molotov.
I hope this tool will be useful. The repo is here: https://github.com/tarekziade/salvo
Happy breaking!