From fcc0d3cb73add2e48f2919db6b51d47c2983bb0c Mon Sep 17 00:00:00 2001 From: Parth Yadav Date: Wed, 15 Jan 2020 12:51:25 +0530 Subject: Support: Container Metrics collection and Visualization A. Collector: Add cAdvisor Collector This patch adds supports for Container Metrics collection using cAdvisor. The collector: 1. Pushes metrics to InfluxDB storage 2. Stores metrics locally in Result Dir 3. Implements print_results() for metric summary 4. Is configurable from 05_collector.conf B. Docker: Grafana Dashboard This patch also adds JSON file for Grafana Dashboard for Container Metrics Signed-off-by: Parth Yadav Change-Id: I7c39d4e1a7a46d3a499a241e618bd0da7eed53a8 --- tools/collectors/cadvisor/cadvisor.py | 218 ++++++++++++++++++++++++++++++++++ 1 file changed, 218 insertions(+) create mode 100644 tools/collectors/cadvisor/cadvisor.py (limited to 'tools/collectors/cadvisor/cadvisor.py') diff --git a/tools/collectors/cadvisor/cadvisor.py b/tools/collectors/cadvisor/cadvisor.py new file mode 100644 index 00000000..de48cecd --- /dev/null +++ b/tools/collectors/cadvisor/cadvisor.py @@ -0,0 +1,218 @@ +# Copyright 2020 University Of Delhi. +# +# 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. + +""" +Collects container metrics from cAdvisor. +Sends metrics to influxDB and also stores results locally. +""" + +import subprocess +import logging +import os +from collections import OrderedDict + +from tools.collectors.collector import collector +from tools import tasks +from conf import settings + + + +# inherit from collector.Icollector. +class Cadvisor(collector.ICollector): + """A collector of container metrics based on cAdvisor + + It starts cadvisor and collects metrics. + """ + + def __init__(self, results_dir, test_name): + """ + Initialize collection of statistics + """ + self._logger = logging.getLogger(__name__) + self.resultsdir = results_dir + self.testname = test_name + self._pid = 0 + self._results = OrderedDict() + self._log = os.path.join(results_dir, + settings.getValue('LOG_FILE_CADVISOR') + + '_' + test_name + '.log') + self._logfile = 0 + + + def start(self): + """ + Starts collection of statistics by cAdvisor and stores them + into- + 1. The file in directory with test results + 2. InfluxDB result container + """ + + # CMD options for cAdvisor + cmd = ['sudo', '/opt/cadvisor/cadvisor', + '-storage_driver='+settings.getValue('CADVISOR_STORAGE_DRIVER'), + '-storage_driver_host='+settings.getValue('CADVISOR_STORAGE_HOST'), + '-storage_driver_db='+settings.getValue('CADVISOR_DRIVER_DB'), + '-housekeeping_interval=0.5s', + '-storage_driver_buffer_duration=1s' + ] + + self._logfile = open(self._log, 'a') + + self._pid = subprocess.Popen(map(os.path.expanduser, cmd), stdout=self._logfile, bufsize=0) + self._logger.info('Starting cAdvisor') + + + + def stop(self): + """ + Stops collection of metrics by cAdvisor and stores statistic + summary for each monitored container into self._results dictionary + """ + try: + subprocess.check_output(["pidof", "cadvisor"]) + tasks.run_task(['sudo', 'pkill', '--signal', '2', 'cadvisor'], + self._logger, 'Stopping cAdvisor', True) + except subprocess.CalledProcessError: + self._logger.error('Failed to stop cAdvisor, maybe process does not exist') + + + self._logfile.close() + self._logger.info('cAdvisor log available at %s', self._log) + + containers = settings.getValue('CADVISOR_CONTAINERS') + self._results = cadvisor_log_result(self._log, containers) + + + def get_results(self): + """Returns collected statistics. + """ + return self._results + + def print_results(self): + """Logs collected statistics. + """ + for cnt in self._results: + logging.info("Container: %s", cnt) + for (key, value) in self._results[cnt].items(): + + postfix = '' + + if key == 'cpu_cumulative_usage': + key = 'CPU_usage' + value = round(float(value) / 1000000000, 4) + postfix = '%' + + if key in ['memory_usage', 'memory_working_set']: + value = round(float(value) / 1024 / 1024, 4) + postfix = 'MB' + + if key in ['rx_bytes', 'tx_bytes']: + value = round(float(value) / 1024 / 1024, 4) + postfix = 'mBps' + + logging.info(" Statistic: %s Value: %s %s", + str(key), str(value), postfix) + + +def cadvisor_log_result(filename, containers): + """ + Processes cAdvisor logfile and returns average results + + :param filename: Name of cadvisor logfile + :param containers: List of container names + + :returns: Result as average stats of Containers + """ + result = OrderedDict() + previous = OrderedDict() + logfile = open(filename, 'r') + with logfile: + # for every line + for _, line in enumerate(logfile): + # skip lines having root '/' metrics + if line[0:7] == 'cName=/': + continue + + # parse line into OrderedDict + tmp_res = parse_line(line) + + cnt = tmp_res['cName'] + + # skip if cnt is not in container list + if cnt not in containers: + continue + + # add metrics to result + if cnt not in result: + result[cnt] = tmp_res + previous[cnt] = tmp_res + result[cnt]['count'] = 1 + else: + for field in tmp_res: + + if field in ['rx_errors', 'tx_errors', 'memory_usage', 'memory_working_set']: + val = float(tmp_res[field]) + elif field in ['cpu_cumulative_usage', 'rx_bytes', 'tx_bytes']: + val = float(tmp_res[field]) - float(previous[cnt][field]) + else: + # discard remaining fields + try: + result[cnt].pop(field) + except KeyError: + continue + continue + + result[cnt][field] = float(result[cnt][field]) + val + + result[cnt]['count'] += 1 + previous[cnt] = tmp_res + + # calculate average results for containers + result = calculate_average(result) + return result + + +def calculate_average(results): + """ + Calculates average for container stats + """ + for cnt in results: + for field in results[cnt]: + if field != 'count': + val = float(results[cnt][field])/results[cnt]['count'] + results[cnt][field] = '{0:.2f}'.format(val) + + results[cnt].pop('count') + #sort results + results[cnt] = OrderedDict(sorted(results[cnt].items())) + + return results + + +def parse_line(line): + """ + Reads single line from cAdvisor logfile + + :param line: single line as str + + :returns: OrderedDict of line read + """ + tmp_res = OrderedDict() + # split line into array of "key=value" metrics + metrics = line.split() + for metric in metrics: + key, value = metric.split('=') + tmp_res[key] = value + + return tmp_res -- cgit 1.2.3-korg