Replicated states HTTP API might have changed
In order to run stress testing on the prototype replicated state, start a cluster with at least 12 DB servers and 3 coordinators (recommended).
Run chaos.py
and follow the logs in the console.
The files actual.json
and expected.json
represent the actual snapshot gathered form the coordinator and the expected final
state which was reconstructed.
There can be a lot of output produced, hence you might want to redirect everything to a file: python chaos.py &> log.txt
How I generally do it:
./scripts/startLocalCluster -a 1 -c 3 -d 12
python chaos.py -op 5000 &> log.txt
Then follow the logs in another terminal:
tail -f log.txt
When you want to re-run the whole thing, don't forget to shutdown the previous cluster.
For debugging purposes, the contents of the replicated log are dumped into replicated-log.json
.