diff options
author | Tibor Frank <tifrank@cisco.com> | 2019-08-02 08:52:26 +0200 |
---|---|---|
committer | Tibor Frank <tifrank@cisco.com> | 2019-08-02 06:55:14 +0000 |
commit | 7b1f19c0c338f5d8eb65fb45b57bb5de84398762 (patch) | |
tree | 5bd46749b03e6e2693cb3e192900ece7aef19378 /resources/tools/presentation | |
parent | d541b2b7d99651b53bd21ff75cd5fdacf8472a98 (diff) |
Trending: Remove multiprocessing from cpta
Change-Id: If53c5a72abed11fda9b15d043b24027d2eb9d86e
Signed-off-by: Tibor Frank <tifrank@cisco.com>
Diffstat (limited to 'resources/tools/presentation')
-rw-r--r-- | resources/tools/presentation/generator_CPTA.py | 70 | ||||
-rw-r--r-- | resources/tools/presentation/input_data_parser.py | 3 |
2 files changed, 17 insertions, 56 deletions
diff --git a/resources/tools/presentation/generator_CPTA.py b/resources/tools/presentation/generator_CPTA.py index a9c4d472ec..4f722574dd 100644 --- a/resources/tools/presentation/generator_CPTA.py +++ b/resources/tools/presentation/generator_CPTA.py @@ -14,8 +14,6 @@ """Generation of Continuous Performance Trending and Analysis. """ -import multiprocessing -import os import logging import csv import prettytable @@ -27,8 +25,7 @@ from collections import OrderedDict from datetime import datetime from copy import deepcopy -from utils import archive_input_data, execute_command, \ - classify_anomalies, Worker +from utils import archive_input_data, execute_command, classify_anomalies # Command to build the html format of the report @@ -324,7 +321,7 @@ def _generate_all_charts(spec, input_data): :type input_data: InputData """ - def _generate_chart(_, data_q, graph): + def _generate_chart(graph): """Generates the chart. """ @@ -499,13 +496,19 @@ def _generate_all_charts(spec, input_data): except plerr.PlotlyEmptyDataError: logs.append(("WARNING", "No data for the plot. Skipped.")) - data_out = { - "job_name": job_name, - "csv_table": csv_tbl, - "results": res, - "logs": logs - } - data_q.put(data_out) + for level, line in logs: + if level == "INFO": + logging.info(line) + elif level == "ERROR": + logging.error(line) + elif level == "DEBUG": + logging.debug(line) + elif level == "CRITICAL": + logging.critical(line) + elif level == "WARNING": + logging.warning(line) + + return {"job_name": job_name, "csv_table": csv_tbl, "results": res} builds_dict = dict() for job in spec.input["builds"].keys(): @@ -534,26 +537,6 @@ def _generate_all_charts(spec, input_data): testbed ) - work_queue = multiprocessing.JoinableQueue() - manager = multiprocessing.Manager() - data_queue = manager.Queue() - cpus = multiprocessing.cpu_count() - - workers = list() - for cpu in range(cpus): - worker = Worker(work_queue, - data_queue, - _generate_chart) - worker.daemon = True - worker.start() - workers.append(worker) - os.system("taskset -p -c {0} {1} > /dev/null 2>&1". - format(cpu, worker.pid)) - - for chart in spec.cpta["plots"]: - work_queue.put((chart, )) - work_queue.join() - anomaly_classifications = list() # Create the header: @@ -570,31 +553,12 @@ def _generate_all_charts(spec, input_data): header = "Version:," + ",".join(versions) + '\n' csv_tables[job_name].append(header) - while not data_queue.empty(): - result = data_queue.get() + for chart in spec.cpta["plots"]: + result = _generate_chart(chart) anomaly_classifications.extend(result["results"]) csv_tables[result["job_name"]].extend(result["csv_table"]) - for item in result["logs"]: - if item[0] == "INFO": - logging.info(item[1]) - elif item[0] == "ERROR": - logging.error(item[1]) - elif item[0] == "DEBUG": - logging.debug(item[1]) - elif item[0] == "CRITICAL": - logging.critical(item[1]) - elif item[0] == "WARNING": - logging.warning(item[1]) - - del data_queue - - # Terminate all workers - for worker in workers: - worker.terminate() - worker.join() - # Write the tables: for job_name, csv_table in csv_tables.items(): file_name = spec.cpta["output-file"] + "-" + job_name + "-trending" diff --git a/resources/tools/presentation/input_data_parser.py b/resources/tools/presentation/input_data_parser.py index 2c248fb3af..3b3be9f1cd 100644 --- a/resources/tools/presentation/input_data_parser.py +++ b/resources/tools/presentation/input_data_parser.py @@ -19,8 +19,6 @@ - filter the data using tags, """ -import multiprocessing -import os import re import resource import pandas as pd @@ -38,7 +36,6 @@ from json import loads from jumpavg.AvgStdevMetadataFactory import AvgStdevMetadataFactory from input_data_files import download_and_unzip_data_file -from utils import Worker # Separator used in file names |