diff options
Diffstat (limited to 'GPL/tools/trex')
-rw-r--r-- | GPL/tools/trex/trex_server_info.py | 54 | ||||
-rw-r--r-- | GPL/tools/trex/trex_stateless_profile.py | 345 | ||||
-rw-r--r-- | GPL/tools/trex/trex_stateless_stop.py | 104 |
3 files changed, 503 insertions, 0 deletions
diff --git a/GPL/tools/trex/trex_server_info.py b/GPL/tools/trex/trex_server_info.py new file mode 100644 index 0000000000..8423801285 --- /dev/null +++ b/GPL/tools/trex/trex_server_info.py @@ -0,0 +1,54 @@ +#!/usr/bin/python3 + +# Copyright (c) 2020 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""This script uses T-REX stateless API to drive t-rex instance. + +Requirements: +- T-REX: https://github.com/cisco-system-traffic-generator/trex-core + - compiled and running T-REX process (eg. ./t-rex-64 -i) + - trex.stl.api library +- Script must be executed on a node with T-REX instance + +Functionality: +1. Verify the API functionality and get server information + +""" + +import sys + +sys.path.insert( + 0, u"/opt/trex-core-2.73/scripts/automation/trex_control_plane/interactive/" +) +from trex.stl.api import * + + +def main(): + """Check server info and quit.""" + client = STLClient() + try: + # connect to server + client.connect() + + # get server info + print(client.get_server_system_info()) + except STLError as ex_error: + print(ex_error, file=sys.stderr) + sys.exit(1) + finally: + client.disconnect() + + +if __name__ == u"__main__": + main() diff --git a/GPL/tools/trex/trex_stateless_profile.py b/GPL/tools/trex/trex_stateless_profile.py new file mode 100644 index 0000000000..a0de5c0cd3 --- /dev/null +++ b/GPL/tools/trex/trex_stateless_profile.py @@ -0,0 +1,345 @@ +#!/usr/bin/python3 + +# Copyright (c) 2020 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""This module gets a traffic profile together with other parameters, reads +the profile and sends the traffic. At the end, it measures the packet loss and +latency. +""" + +import argparse +import json +import sys +import time + +sys.path.insert( + 0, u"/opt/trex-core-2.73/scripts/automation/trex_control_plane/interactive/" +) +from trex.stl.api import * + + +def fmt_latency(lat_min, lat_avg, lat_max, hdrh): + """Return formatted, rounded latency. + + :param lat_min: Min latency + :param lat_avg: Average latency + :param lat_max: Max latency + :param hdrh: Base64 encoded compressed HDRHistogram object. + :type lat_min: str + :type lat_avg: str + :type lat_max: str + :type hdrh: str + :return: Formatted and rounded output (hdrh unchanged) "min/avg/max/hdrh". + :rtype: str + """ + try: + t_min = int(round(float(lat_min))) + except ValueError: + t_min = int(-1) + try: + t_avg = int(round(float(lat_avg))) + except ValueError: + t_avg = int(-1) + try: + t_max = int(round(float(lat_max))) + except ValueError: + t_max = int(-1) + + return u"/".join(str(tmp) for tmp in (t_min, t_avg, t_max, hdrh)) + + +def simple_burst( + profile_file, duration, framesize, rate, warmup_time, port_0, port_1, + latency, async_start=False, traffic_directions=2, force=False): + """Send traffic and measure packet loss and latency. + + Procedure: + - reads the given traffic profile with streams, + - connects to the T-rex client, + - resets the ports, + - removes all existing streams, + - adds streams from the traffic profile to the ports, + - if the warm-up time is more than 0, sends the warm-up traffic, reads the + statistics, + - clears the statistics from the client, + - starts the traffic, + - waits for the defined time (or runs forever if async mode is defined), + - stops the traffic, + - reads and displays the statistics and + - disconnects from the client. + + :param profile_file: A python module with T-rex traffic profile. + :param framesize: Frame size. + :param duration: Duration of traffic run in seconds (-1=infinite). + :param rate: Traffic rate [percentage, pps, bps]. + :param warmup_time: Traffic warm-up time in seconds, 0 = disable. + :param port_0: Port 0 on the traffic generator. + :param port_1: Port 1 on the traffic generator. + :param latency: With latency stats. + :param async_start: Start the traffic and exit. + :param traffic_directions: Bidirectional (2) or unidirectional (1) traffic. + :param force: Force start regardless of ports state. + :type profile_file: str + :type framesize: int or str + :type duration: float + :type rate: str + :type warmup_time: float + :type port_0: int + :type port_1: int + :type latency: bool + :type async_start: bool + :type traffic_directions: int + :type force: bool + """ + client = None + total_rcvd = 0 + total_sent = 0 + approximated_duration = 0 + approximated_rate = 0 + lost_a = 0 + lost_b = 0 + lat_a = u"-1/-1/-1/" + lat_b = u"-1/-1/-1/" + + # Read the profile: + try: + print(f"### Profile file:\n{profile_file}") + profile = STLProfile.load( + profile_file, direction=0, port_id=0, framesize=framesize + ) + streams = profile.get_streams() + except STLError as err: + print(f"Error while loading profile '{profile_file}' {err!r}") + sys.exit(1) + + try: + # Create the client: + client = STLClient() + # Connect to server: + client.connect() + # Prepare our ports (the machine has 0 <--> 1 with static route): + client.reset(ports=[port_0, port_1]) + client.remove_all_streams(ports=[port_0, port_1]) + + if u"macsrc" in profile_file: + client.set_port_attr(ports=[port_0, port_1], promiscuous=True) + if isinstance(framesize, int): + client.add_streams(streams[0], ports=[port_0]) + if traffic_directions > 1: + client.add_streams(streams[1], ports=[port_1]) + elif isinstance(framesize, str): + client.add_streams(streams[0:3], ports=[port_0]) + if traffic_directions > 1: + client.add_streams(streams[3:6], ports=[port_1]) + if latency: + try: + if isinstance(framesize, int): + client.add_streams(streams[2], ports=[port_0]) + if traffic_directions > 1: + client.add_streams(streams[3], ports=[port_1]) + elif isinstance(framesize, str): + latency = False + except STLError: + # Disable latency if NIC does not support requested stream type + print(u"##### FAILED to add latency streams #####") + latency = False + ports = [port_0] + if traffic_directions > 1: + ports.append(port_1) + # Warm-up phase: + if warmup_time > 0: + # Clear the stats before injecting: + client.clear_stats() + + # Choose rate and start traffic: + client.start(ports=ports, mult=rate, duration=warmup_time, + force=force) + + # Block until done: + time_start = time.monotonic() + client.wait_on_traffic(ports=ports, timeout=warmup_time+30) + time_stop = time.monotonic() + approximated_duration = time_stop - time_start + + if client.get_warnings(): + for warning in client.get_warnings(): + print(warning) + + # Read the stats after the test: + stats = client.get_stats() + + print(u"##### Warmup statistics #####") + print(json.dumps(stats, indent=4, separators=(u",", u": "))) + + lost_a = stats[port_0][u"opackets"] - stats[port_1][u"ipackets"] + if traffic_directions > 1: + lost_b = stats[port_1][u"opackets"] - stats[port_0][u"ipackets"] + + print(f"\npackets lost from {port_0} --> {port_1}: {lost_a} pkts") + if traffic_directions > 1: + print(f"packets lost from {port_1} --> {port_0}: {lost_b} pkts") + + # Clear the stats before injecting: + client.clear_stats() + lost_a = 0 + lost_b = 0 + + # Choose rate and start traffic: + client.start(ports=ports, mult=rate, duration=duration) + + if async_start: + # For async stop, we need to export the current snapshot. + xsnap0 = client.ports[0].get_xstats().reference_stats + print(f"Xstats snapshot 0: {xsnap0!r}") + if traffic_directions > 1: + xsnap1 = client.ports[1].get_xstats().reference_stats + print(f"Xstats snapshot 1: {xsnap1!r}") + else: + # Block until done: + time_start = time.monotonic() + client.wait_on_traffic(ports=ports, timeout=duration+30) + time_stop = time.monotonic() + approximated_duration = time_stop - time_start + + if client.get_warnings(): + for warning in client.get_warnings(): + print(warning) + + # Read the stats after the test + stats = client.get_stats() + + print(u"##### Statistics #####") + print(json.dumps(stats, indent=4, separators=(u",", u": "))) + + lost_a = stats[port_0][u"opackets"] - stats[port_1][u"ipackets"] + if traffic_directions > 1: + lost_b = stats[port_1][u"opackets"] - stats[port_0][u"ipackets"] + + # Stats index is not a port number, but "pgid". + if latency: + lat_obj = stats[u"latency"][0][u"latency"] + lat_a = fmt_latency( + str(lat_obj[u"total_min"]), str(lat_obj[u"average"]), + str(lat_obj[u"total_max"]), str(lat_obj[u"hdrh"])) + if traffic_directions > 1: + lat_obj = stats[u"latency"][1][u"latency"] + lat_b = fmt_latency( + str(lat_obj[u"total_min"]), str(lat_obj[u"average"]), + str(lat_obj[u"total_max"]), str(lat_obj[u"hdrh"])) + + if traffic_directions > 1: + total_sent = stats[0][u"opackets"] + stats[1][u"opackets"] + total_rcvd = stats[0][u"ipackets"] + stats[1][u"ipackets"] + else: + total_sent = stats[port_0][u"opackets"] + total_rcvd = stats[port_1][u"ipackets"] + try: + approximated_rate = total_sent / approximated_duration + except ZeroDivisionError: + pass + + print(f"\npackets lost from {port_0} --> {port_1}: {lost_a} pkts") + if traffic_directions > 1: + print(f"packets lost from {port_1} --> {port_0}: {lost_b} pkts") + + except STLError as ex_error: + print(ex_error, file=sys.stderr) + sys.exit(1) + + finally: + if async_start: + if client: + client.disconnect(stop_traffic=False, release_ports=True) + else: + if client: + client.disconnect() + print( + f"rate={rate!r}, totalReceived={total_rcvd}, " + f"totalSent={total_sent}, frameLoss={lost_a + lost_b}, " + f"targetDuration={duration!r}, " + f"approximatedDuration={approximated_duration!r}, " + f"approximatedRate={approximated_rate}, " + f"latencyStream0(usec)={lat_a}, latencyStream1(usec)={lat_b}, " + ) + + +def main(): + """Main function for the traffic generator using T-rex. + + It verifies the given command line arguments and runs "simple_burst" + function. + """ + parser = argparse.ArgumentParser() + parser.add_argument( + u"-p", u"--profile", required=True, type=str, + help=u"Python traffic profile." + ) + parser.add_argument( + u"-d", u"--duration", required=True, type=float, + help=u"Duration of traffic run." + ) + parser.add_argument( + u"-s", u"--frame_size", required=True, + help=u"Size of a Frame without padding and IPG." + ) + parser.add_argument( + u"-r", u"--rate", required=True, + help=u"Traffic rate with included units (%, pps)." + ) + parser.add_argument( + u"-w", u"--warmup_time", type=float, default=5.0, + help=u"Traffic warm-up time in seconds, 0 = disable." + ) + parser.add_argument( + u"--port_0", required=True, type=int, + help=u"Port 0 on the traffic generator." + ) + parser.add_argument( + u"--port_1", required=True, type=int, + help=u"Port 1 on the traffic generator." + ) + parser.add_argument( + u"--async_start", action=u"store_true", default=False, + help=u"Non-blocking call of the script." + ) + parser.add_argument( + u"--latency", action=u"store_true", default=False, + help=u"Add latency stream." + ) + parser.add_argument( + u"--traffic_directions", type=int, default=2, + help=u"Send bi- (2) or uni- (1) directional traffic." + ) + parser.add_argument( + u"--force", action=u"store_true", default=False, + help=u"Force start regardless of ports state." + ) + + args = parser.parse_args() + + try: + framesize = int(args.frame_size) + except ValueError: + framesize = args.frame_size + + simple_burst( + profile_file=args.profile, duration=args.duration, framesize=framesize, + rate=args.rate, warmup_time=args.warmup_time, port_0=args.port_0, + port_1=args.port_1, latency=args.latency, async_start=args.async_start, + traffic_directions=args.traffic_directions, force=args.force + ) + + +if __name__ == u"__main__": + main() diff --git a/GPL/tools/trex/trex_stateless_stop.py b/GPL/tools/trex/trex_stateless_stop.py new file mode 100644 index 0000000000..5e8721d28c --- /dev/null +++ b/GPL/tools/trex/trex_stateless_stop.py @@ -0,0 +1,104 @@ +#!/usr/bin/python3 + +# Copyright (c) 2020 Cisco and/or its affiliates. +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at: +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +"""This script uses T-REX stateless API to drive t-rex instance. + +Requirements: +- T-REX: https://github.com/cisco-system-traffic-generator/trex-core + - compiled and running T-REX process (eg. ./t-rex-64 -i) + - trex.stl.api library +- Script must be executed on a node with T-REX instance + +Functionality: +1. Stop any running traffic +2. Optionally restore reference counter values. +3. Return conter differences. +""" + +import argparse +import json +import sys + +from collections import OrderedDict # Needed to parse xstats representation. + +sys.path.insert( + 0, u"/opt/trex-core-2.73/scripts/automation/trex_control_plane/interactive/" +) +from trex.stl.api import * + + +def main(): + """Stop traffic if any is running. Report xstats.""" + parser = argparse.ArgumentParser() + parser.add_argument( + u"--xstat0", type=str, default=u"", + help=u"Reference xstat object if any." + ) + parser.add_argument( + u"--xstat1", type=str, default=u"", + help=u"Reference xstat object if any." + ) + args = parser.parse_args() + + client = STLClient() + try: + # connect to server + client.connect() + + client.acquire(force=True) + # TODO: Support unidirection. + client.stop(ports=[0, 1]) + + # Read the stats after the test, + # we need to update values before the last trial started. + if args.xstat0: + snapshot = eval(args.xstat0) + client.ports[0].get_xstats().reference_stats = snapshot + if args.xstat1: + snapshot = eval(args.xstat1) + client.ports[1].get_xstats().reference_stats = snapshot + # Now we can call the official method to get differences. + xstats0 = client.get_xstats(0) + xstats1 = client.get_xstats(1) + + # If STLError happens, let the script fail with stack trace. + finally: + client.disconnect() + + print(u"##### statistics port 0 #####") + print(json.dumps(xstats0, indent=4, separators=(u",", u": "))) + print(u"##### statistics port 1 #####") + print(json.dumps(xstats1, indent=4, separators=(u",", u": "))) + + tx_0, rx_0 = xstats0[u"tx_good_packets"], xstats0[u"rx_good_packets"] + tx_1, rx_1 = xstats1[u"tx_good_packets"], xstats1[u"rx_good_packets"] + lost_a, lost_b = tx_0 - rx_1, tx_1 - rx_0 + + print(f"\npackets lost from 0 --> 1: {lost_a} pkts") + print(f"packets lost from 1 --> 0: {lost_b} pkts") + + total_rcvd, total_sent = rx_0 + rx_1, tx_0 + tx_1 + total_lost = total_sent - total_rcvd + # TODO: Add latency. + print( + f"rate='unknown', totalReceived={total_rcvd}, totalSent={total_sent}, " + f"frameLoss={total_lost}, targetDuration='manual', " + f"approximatedDuration='manual', approximatedRate='unknown', " + f"latencyStream0(usec)=-1/-1/-1, latencyStream1(usec)=-1/-1/-1" + ) + + +if __name__ == u"__main__": + main() |