Source code for perfmon.core.metrics.init

"""This file contains class to launch monitoring process"""

import os
import logging
import time
import copy
import multiprocessing
import threading
import psutil

from perfmon.common.pid import GetJobPid
from perfmon.common.utils.json_wrappers import dump_json
from perfmon.core.metrics.cpu import MonitorCpuUsage
from perfmon.core.metrics.perfcounters import MonitorPerfCounters
from perfmon.core.metrics.gpu import MonitorNvidiaGpuMetrics

_log = logging.getLogger(__name__)

# pylint: disable=E0401,W0201,C0301


[docs]class MonitorPerformanceMetrics(object):
    """Engine to extract performance metrics"""

    # pylint: disable=too-many-instance-attributes

    def __init__(self, config):
        """Initialize setup"""

        self.config = config
        self._extra = {}

[docs]    def get_job_pid(self):
        """This method calls function to get job PID"""

        # Get pid of the main job to monitor
        pid_getter = GetJobPid(self.config)
        self.config['pid'] = pid_getter.go()

        _log.info('The monitoring process has the PID: %d', os.getpid())
        _log.info(
            'The job to monitor has PID(s): %s', ','.join([str(p) for p in self.config['pid']])
        )

[docs]    def start_collection(self):
        """Start collecting CPU metrics. We use multiprocessing library to spawn different
        processes to monitor cpu and perf metrics"""

        self._extra['start_time'] = time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())

        # Find job pid
        self.get_job_pid()

        # Spawn processes to monitor the job simultaneously.
        metric_processes = {}
        if 'cpu_metrics' in self.config['metrics']:
            metric_processes['cpu_metrics'] = MonitorCpuUsage(config=self.config)
            metric_processes['cpu_metrics'].start()

        if 'perf_metrics' in self.config['metrics']:
            metric_processes['perf_metrics'] = MonitorPerfCounters(config=self.config)
            metric_processes['perf_metrics'].start()

        if 'nv_gpu_metrics' in self.config['metrics']:
            metric_processes['nvidia_gpu'] = MonitorNvidiaGpuMetrics(config=self.config)
            metric_processes['nvidia_gpu'].start()

        # Wait for the spawned processes to finish
        for _, spawned_proc in metric_processes.items():
            spawned_proc.join()

        self._extra['end_time'] = time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())
Source code for perfmon.core.metrics.__init__

Source code for perfmon.core.metrics.init
