A docker benchmarking framework for Kong that runs a series of load tests to determine performance under various conditions such as with plugins enabled and a large number of concurrent connections.
The benchmarks are performed by Siege and results are written to /usr/local/kong/siege.log
so you can just add -v /local/path:/usr/local/kong
when running a container to expose it as a persistent volume containing the log file. If you're running boot2docker you'll have to use a local path inside the /Users
directory.
If you format Siege's log it's possible to get a nice CSV file that you can chart.
# build it
git clone [email protected]:montanaflynn/kong-benchmarks.git
cd kong-benchmarks
docker build -t "kong-benchmarks" .
# run it
docker run kong-benchmarks
There are some parameters you can modify by passing environment variables to docker run
:
name | default | description |
---|---|---|
TIME | "60S" | Time to siege for |
UPSTREAM | "http://127.0.0.1:8001/robots.txt" | Upstream target to siege |
VERBOSE_LOGS | false | Maximum concurrency |
DEBUG | false | Maximum concurrency |
WARMUP | true | Benchmark Kong plugins |
PLUGINS | true | Benchmark Kong plugins |
RUN_CASSANDRA | true | Run Cassandra |
CONCURRENT | 100 | Starting concurrency |
MAX_CONCURRENT | 500 | Maximum concurrency |
Use external Cassandra by replacing the kong.yml
file and not running Cassandra:
docker run -e "RUN_CASSANDRA=false" -v $PWD/kong-config/:/etc/kong/ kong-benchmarks
Using mockbin as the upstream API up to 100 concurrent connections in increments of 20:
docker run -e "CONCURRENT=20" -e "MAX_CUNCURRENT=100" -e "UPSTREAM=http://mockbin.com/status/418" kong-benchmarks
Here's how you can run a short test of 50 concurrent connections without plugins:
docker run -e "CONCURRENT=50" -e "MAX_CUNCURRENT=50" -e "PLUGINS=false" kong-benchmarks
Have a persistent log file on your docker host, subsequent runs will append to the same file:
docker run -v /logs/kong-benchmarks/:/usr/local/kong kong-benchmarks