This scenario attempts to generate a reasonable amount of load by hitting the
HTTP interface of the deployed Tendermint nodes, assuming that the kvstore
proxy app is running.
To execute the load test from a single machine:
# Optional: deploy a clean reference network
make deploy:001-reference
# Standard load test (1000 clients, spawning 20 clients/sec, running for 60s)
# By default, this hits the following URLs:
# - http://tik0.sredev.co:26657
# - http://tik1.sredev.co:26657
# - http://tik2.sredev.co:26657
# - http://tik3.sredev.co:26657
make scenario:002-kvstore-loadtest
# Custom parameters for load test.
# This eventually spawns up to 2000 clients, spawning at 40 clients/sec,
# running for a total time of 10 minutes.
# Note that host URLs are separated by double colons (::).
HOST_URLS=http://tik2.sredev.co:26657::http://tik3.sredev.co:26657 \
NUM_CLIENTS=2000 \
HATCH_RATE=40 \
RUN_TIME=10m \
make scenario:002-kvstore-loadtest
Load statistics will be shown on stdout
, and the following files will be
generated:
/tmp/002-kvstore-loadtest.log
- The log file detailing specific errors or issues that occurred during load testing./tmp/002-kvstore-loadtest_distribution.csv
- A distribution of the response times for each class of query./tmp/002-kvstore-loadtest_requests.csv
- Request statistics for each class of query.