The idea is to be able to specify size/type of Riak cluster, and size/type of benchmark, and this will automatically spin up AWS instances, benchmark, record info, save all the results, and then blow away all nodes.
- creates a VPC
- attaches/formats/configures volumes of custom type
- tunes instances
- uses influxdb and configures telegraf on all nodes
- imports all riak stats using telegraf
- uses distributed basho_bench & generates summary
- copies influxdb dataset & benchmark results locally
- wipes everything after benchmark
-
set your boto/AWS creds:
-
copy and edit
example-bb.json
options -
run
./start_bench.sh
- by default, start_bench.sh will also wipe the cluster once it's done
- doesn't install grafana by default. json dashboard for convenience
- test local influxdb data recovery
- add option to request spot instances
- create influxdb/grafana docker to view influxdb data
- option to use local/persistant influxdb