Devtools hangs

Devtools hangs.ipynb

Devtools hangs.py

# coding: utf-8

# In[1]:

import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
from datetime import datetime, timedelta
from moztelemetry import get_pings_properties
from moztelemetry.dataset import Dataset
from scipy.stats import linregress
from sets import Set

sc.defaultParallelism


# In[2]:

bhr_pings = (Dataset.from_source("telemetry")
    .where(docType='OTHER')
    .where(appVersion=lambda v: v >= '57')
    .where(appUpdateChannel="nightly")
    .records(sc, sample=1.0))

bhr_pings = bhr_pings.filter(lambda p: p.get('meta', {}).get('docType', {}) == 'bhr')

bhr_properties = [
    "meta/docType",
    "environment/settings/e10sEnabled",
    "environment/system/os/name",
    "environment/system/os/version",
    "application/architecture",
    "application/buildId",
    "payload/modules",
    "payload/hangs",
    "payload/timeSinceLastPing",
]
bhr_subset = get_pings_properties(bhr_pings, bhr_properties)


# In[18]:

def flatten_hangs(thread_hang):
    if 'name' not in thread_hang:
        return []

    hangs = thread_hang['hangs']

    return [
        {
            'thread_name': thread_hang['name'],
            'hang': x
        }
        for x in hangs
    ]

def flatten_all_hangs(ping):
    return ping['payload/hangs']

def has_devtools_stack(hang):
    return any("devtools" in s for s in hang["stack"])

def count_bhr_hangs(thread_name, hangs, min_duration, max_duration):
    count = 0
    for hang in hangs:
        if (hang['thread'] == thread_name and
            hang['duration'] >= min_duration and
            hang['duration'] <= max_duration and
            has_devtools_stack(hang)):
            count += 1
    return count

def map_ping(ping):
    hangs = flatten_all_hangs(ping)
    
    subsession_length = ping['payload/timeSinceLastPing']
    return (ping["application/buildId"][:8], {
        'subsession_length': subsession_length,
        'parent_bhr_512': count_bhr_hangs('Gecko', hangs, 512, 2048),
        'parent_bhr_2048': count_bhr_hangs('Gecko', hangs, 2048, 65538),
        'content_bhr_512': count_bhr_hangs('Gecko_Child', hangs, 512, 2048),
        'content_bhr_2048': count_bhr_hangs('Gecko_Child', hangs, 2048, 65538),
    })

def merge_counts(a, b):
    return {k: a[k] + b[k] for k in a.iterkeys()}

def ping_is_valid(ping):
    if type(ping["payload/timeSinceLastPing"]) != int:
        return False
    
    if not isinstance(ping["application/buildId"], basestring):
        return False
    
    if ping["application/architecture"] != "x86-64":
        return False
    
    if not ping["environment/system/os/version"].startswith("10"):
        return False
    
    if not ping["environment/settings/e10sEnabled"]:
        return False

    return ping["environment/system/os/name"] == "Windows_NT"

counts = bhr_subset.cache().filter(ping_is_valid).map(map_ping).reduceByKey(merge_counts).collectAsMap()


# In[28]:

def moving_average(a, n=3) :
    ret = np.cumsum(a, dtype=float)
    ret[n:] = ret[n:] - ret[:-n]
    return ret / n


# In[25]:

import matplotlib.pyplot as plt
import numpy as np
get_ipython().magic(u'matplotlib inline')

data = counts
sorted_keys = sorted(data.iterkeys())
plot_data_2048 = np.array([float(data[k]['parent_bhr_2048']) / data[k]['subsession_length'] * 3600. * 1000 for k in sorted_keys], np.float32)
plot_data_2048 = moving_average(plot_data_2048, 7)
plot_data_512 = np.array([float(data[k]['parent_bhr_512']) / data[k]['subsession_length'] * 3600. * 1000 for k in sorted_keys], np.float32)
plot_data_512 = moving_average(plot_data_512, 7)

plt.title("Devtools Hang Stats")
plt.xlabel("Build date")
plt.ylabel("Hangs per 1000 usage hours")

plt.xticks(range(0, len(data)))
max_y = np.amax(plot_data_512)
plt.yticks(np.arange(0., max_y, max_y / 20.))

xticks = np.arange(0, len(plot_data_2048), len(plot_data_2048) / 6)

plt.xticks(xticks, (sorted_keys[int(i)][4:8] for i in xticks))

x1,x2,y1,y2 = plt.axis()
plt.axis((x1,x2,0,max_y))

plt.plot(range(0, len(data)), plot_data_512)
plt.plot(range(0, len(data)), plot_data_2048)

plt.legend(["2048ms > x > 512ms", "x > 2048ms"], loc="upper left")


# In[27]:

import matplotlib.pyplot as plt
import numpy as np
get_ipython().magic(u'matplotlib inline')

data = counts
sorted_keys = sorted(data.iterkeys())
plot_data_2048 = np.array([float(data[k]['content_bhr_2048']) / data[k]['subsession_length'] * 3600. * 1000 for k in sorted_keys], np.float32)
plot_data_2048 = moving_average(plot_data_2048, 7)
plot_data_512 = np.array([float(data[k]['content_bhr_512']) / data[k]['subsession_length'] * 3600. * 1000 for k in sorted_keys], np.float32)
plot_data_512 = moving_average(plot_data_512, 7)

plt.title("Content devtools Hang Stats")
plt.xlabel("Build date")
plt.ylabel("Hangs per 1000 usage hours")

plt.xticks(range(0, len(data)))
max_y = np.amax(plot_data_512)
plt.yticks(np.arange(0., max_y, max_y / 20.))

xticks = np.arange(0, len(plot_data_2048), len(plot_data_2048) / 6)

plt.xticks(xticks, (sorted_keys[int(i)][4:8] for i in xticks))

x1,x2,y1,y2 = plt.axis()
plt.axis((x1,x2,0,max_y))

plt.plot(range(0, len(data)), plot_data_2048)
plt.plot(range(0, len(data)), plot_data_512)

plt.legend(["2048ms > x > 512ms", "x > 2048ms"], loc="upper left")