The MZBench CLI lets you control you control the server and benchmarks from the command line. It utilizes the MZBench API, but goes beyond it: it can do things even without a running MZBench server.

The commands are invoked by the mzbench script in the bin directory:

# Inside the MZBench directory:
$ ./bin/mzbench start --env foo=bar --nodes=5
# or:
$ cd bin
$ mzbench start --env foo=bar --nodes=5

Commands

Server Control

start_server

$ ./bin/mzbench start_server
Executing make -C /path/to/mzbench/bin/../server generate
...

Start the MZBench server.

Optional params:

--config <config_file> : Path to the server config file.

stop_server

$ ./bin/mzbench stop_server
Executing make -C /path/to/mzbench/bin/../server generate
...

Stop the MZBench server.

restart_server

$ ./bin/mzbench restart_server
Executing make -C /path/to/mzbench/bin/../server generate
...

Restart the MZBench server, i.e. stop + start.

Optional params:

--config <config_file> : Path to the server config file.

Benchmark Control

start

$ mzbench start --env foo=bar --nodes=5 foo.erl
{
    "status": "pending", 
    "id": 86
}

Start the benchmark from the given scenario file.

Positional param:

<scenario_file> : The path to the scenario file for the benchmark.

Optional params:

--name <benchmark_name> : Benchmark name.

--nodes <nodes> : Number of nodes or a comma-separated list of node hostnames to run the benchmark on.

--nodes_file <filename> : Path to a file with node hostnames separated by newlines.

--env <name=value> ... : Environment variable definitions.

--cloud <cloud_provider_name> : Name of the cloud provider from the server config. If not specified, the first one on the list is used.

--email <email> ... : Emails for notifications. When the benchmark finishes, the results will be sent to these emails.

--deallocate_after_bench false : Skip node deallocation after the benchmark.

--provision_nodes false : Skip MZBench installation on the nodes.

--exclusive_node_usage false : Allow multiple nodes to be hosted on the same physical machine, i.e. do not allocate a physical machine exclusively for each node.

--node_commit=<commit> : Commit hash or branch name in the MZBench repository pointing to the MZBench version to install on the nodes.

run

$ mzbench run --env foo=bar --nodes=5 foo.erl
{
    "status": "pending", 
    "id": 86
}

Same as start, but blocks until the benchmark is complete.

Positional param:

<scenario_file> : The path to the scenario file for the benchmark.

run_local

$ mzbench run_local --env foo=bar foo.erl
Executing make -C /path/to/mzbench/bin/../node compile
...

Run the benchmark without a server. The logs are printed to stdout.

Positional param:

<scenario_file> : The path to the scenario file for the benchmark.

Optional param:

--env <name=value> ... : Environment variable definitions.

validate

$ mzbench validate foo.erl
ok

Validate the scenario file without executing it.

Positional param:

<scenario_file> : The path to the scenario file for the benchmark.