""" yardstick - command line tool for managing benchmarks
Example invocation:
- $ yardstick samples/ping-task.yaml
-"""
-
-import sys
-import yaml
-import atexit
-import pkg_resources
-
-from yardstick.benchmark.context.model import Context
-from yardstick.benchmark.runners import base as base_runner
-from yardstick.cmdparser import CmdParser
-from yardstick.orchestrator.heat import HeatStack
-
-
-class TaskParser(object):
- '''Parser for task config files in yaml format'''
- def __init__(self, path):
- self.path = path
-
- def parse(self):
- '''parses the task file and return an context and scenario instances'''
- print "Parsing task config:", self.path
- try:
- with open(self.path) as stream:
- cfg = yaml.load(stream)
- except IOError as ioerror:
- sys.exit(ioerror)
-
- if cfg["schema"] != "yardstick:task:0.1":
- sys.exit("error: file %s has unknown schema %s" % (self.path,
- cfg["schema"]))
- context = Context()
- context.init(cfg["context"])
-
- run_in_parallel = cfg.get("run_in_parallel", False)
-
- # TODO we need something better here, a class that represent the file
- return cfg["scenarios"], run_in_parallel, context
-
-
-def atexit_handler():
- '''handler for process termination'''
- if HeatStack.stacks_exist():
- print "Deleting all stacks"
- HeatStack.delete_all()
+ $ yardstick task start samples/ping.yaml
+ Servers are the same as VMs (Nova calls them servers in the API)
-def run_one_scenario(scenario_cfg, context, output_file):
- '''run one scenario using context'''
- key_filename = pkg_resources.resource_filename(
- 'yardstick.resources', 'files/yardstick_key')
+ Many tests use a client/server architecture. A test client is configured
+ to use a specific test server e.g. using an IP address. This is true for
+ example iperf. In some cases the test server is included in the kernel
+ (ping, pktgen) and no additional software is needed on the server. In other
+ cases (iperf) a server process needs to be installed and started.
- host = context.get_server(scenario_cfg["host"])
+ One server is required to host the test client program (such as ping or
+ iperf). In the task file this server is called host.
- runner_cfg = scenario_cfg["runner"]
- runner_cfg['host'] = host.floating_ip["ipaddr"]
- runner_cfg['user'] = context.user
- runner_cfg['key_filename'] = key_filename
- runner_cfg['output_filename'] = output_file
+ A server can be the _target_ of a test client (think ping destination
+ argument). A target server is optional but needed in most test scenarios.
+ In the task file this server is called target. This is probably the same
+ as DUT in existing terminology.
- target = context.get_server(scenario_cfg["target"])
- if target.floating_ip:
- runner_cfg['target'] = target.floating_ip["ipaddr"]
+ Existing terminology:
+ https://www.ietf.org/rfc/rfc1242.txt (throughput/latency)
+ https://www.ietf.org/rfc/rfc2285.txt (DUT/SUT)
- # TODO hardcoded name below, a server can be attached to several nets
- scenario_cfg["ipaddr"] = target.ports["test"]["ipaddr"]
+ New terminology:
+ NFV TST
- runner = base_runner.Runner.get(runner_cfg)
-
- print "Starting runner of type '%s'" % runner_cfg["type"]
- runner.run(scenario_cfg["type"], scenario_cfg)
+"""
+import sys
- return runner
+from yardstick.cmd.cli import YardstickCLI
def main():
'''yardstick main'''
-
- atexit.register(atexit_handler)
-
- prog_args = CmdParser().parse_args()
-
- parser = TaskParser(prog_args.taskfile[0])
- scenarios, run_in_parallel, context = parser.parse()
-
- if prog_args.parse_only:
- sys.exit(0)
-
- context.deploy()
-
- runners = []
- if run_in_parallel:
- for scenario in scenarios:
- runner = run_one_scenario(scenario, context, prog_args.output_file)
- runners.append(runner)
-
- # Wait for runners to finish
- for runner in runners:
- runner.join()
- print "Runner ended, output in", prog_args.output_file
- base_runner.Runner.release(runner)
- else:
- # run serially
- for scenario in scenarios:
- runner = run_one_scenario(scenario, context, prog_args.output_file)
- runner.join()
- print "Runner ended, output in", prog_args.output_file
- base_runner.Runner.release(runner)
-
- if prog_args.keep_deploy:
- # keep deployment, forget about stack (hide it for exit handler)
- context.stack = None
- else:
- context.undeploy()
-
- print "Done, exiting"
+ YardstickCLI().main(sys.argv[1:])
if __name__ == '__main__':
main()
Code Review / yardstick.git / blobdiff