##// END OF EJS Templates
worker: change partition strategy to every Nth element...
worker: change partition strategy to every Nth element The only consumer of the worker pool code today is `hg update`. Previously, the algorithm to partition work to each worker process preserved input list ordering. We'd take the first N elements, then the next N elements, etc. Measurements on mozilla-central demonstrate this isn't an optimal partitioning strategy. I added debug code to print when workers were exiting. When performing a working copy update on a previously empty working copy of mozilla-central, I noticed that process lifetimes were all over the map. One worker would complete after 7s. Many would complete after 12s. And another worker would often take >16s. This behavior occurred for many worker process counts and was more pronounced on some than others. What I suspect is happening is some workers end up with lots of small files and others with large files. This is because the update code passes in actions according to sorted filenames. And, directories under tend to accumulate similar files. For example, test directories often consist of many small test files and media directories contain binary (often larger) media files. This patch changes the partitioning algorithm to select every Nth element from the input list. Each worker thus has a similar composition of files to operate on. The result of this change is that worker processes now all tend to exit around the same time. The possibility of a long pole due to being unlucky and receiving all the large files has been mitigated. Overall execution time seems to drop, but not by a statistically significant amount on mozilla-central. However, repositories with directories containing many large files will likely show a drop. There shouldn't be any regressions due to partial manifest decoding because the update code already iterates the manifest to determine what files to operate on, so the manifest should already be decoded.

File last commit:

r28181:f8efc8a3 default
r28181:f8efc8a3 default
Show More
worker.py
162 lines | 4.5 KiB | text/x-python | PythonLexer
Bryan O'Sullivan
worker: count the number of CPUs...
r18635 # worker.py - master-slave parallelism support
#
# Copyright 2013 Facebook, Inc.
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
Gregory Szorc
worker: use absolute_import
r25992 from __future__ import absolute_import
import errno
import os
import signal
import sys
import threading
from .i18n import _
Pierre-Yves David
error: get Abort from 'error' instead of 'util'...
r26587 from . import error
Bryan O'Sullivan
worker: count the number of CPUs...
r18635
def countcpus():
'''try to count the number of CPUs on the system'''
Gregory Szorc
worker: restore old countcpus code (issue4869)...
r26568
# posix
Bryan O'Sullivan
worker: count the number of CPUs...
r18635 try:
Gregory Szorc
worker: restore old countcpus code (issue4869)...
r26568 n = int(os.sysconf('SC_NPROCESSORS_ONLN'))
if n > 0:
return n
except (AttributeError, ValueError):
pass
# windows
try:
n = int(os.environ['NUMBER_OF_PROCESSORS'])
if n > 0:
return n
except (KeyError, ValueError):
pass
return 1
Bryan O'Sullivan
worker: estimate whether it's worth running a task in parallel...
r18636
def _numworkers(ui):
s = ui.config('worker', 'numcpus')
if s:
try:
n = int(s)
if n >= 1:
return n
except ValueError:
Pierre-Yves David
error: get Abort from 'error' instead of 'util'...
r26587 raise error.Abort(_('number of cpus must be an integer'))
Bryan O'Sullivan
worker: estimate whether it's worth running a task in parallel...
r18636 return min(max(countcpus(), 4), 32)
if os.name == 'posix':
_startupcost = 0.01
else:
_startupcost = 1e30
def worthwhile(ui, costperop, nops):
'''try to determine whether the benefit of multiple processes can
outweigh the cost of starting them'''
linear = costperop * nops
workers = _numworkers(ui)
benefit = linear - (_startupcost * workers + linear / workers)
return benefit >= 0.15
Bryan O'Sullivan
worker: partition a list (of tasks) into equal-sized chunks
r18637
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 def worker(ui, costperarg, func, staticargs, args):
'''run a function, possibly in parallel in multiple worker
processes.
returns a progress iterator
costperarg - cost of a single task
func - function to run
staticargs - arguments to pass to every invocation of the function
args - arguments to split into chunks, to pass to individual
workers
'''
if worthwhile(ui, costperarg, len(args)):
return _platformworker(ui, func, staticargs, args)
return func(*staticargs + (args,))
def _posixworker(ui, func, staticargs, args):
rfd, wfd = os.pipe()
workers = _numworkers(ui)
Bryan O'Sullivan
worker: fix a race in SIGINT handling...
r18708 oldhandler = signal.getsignal(signal.SIGINT)
signal.signal(signal.SIGINT, signal.SIG_IGN)
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 pids, problem = [], [0]
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 for pargs in partition(args, workers):
pid = os.fork()
if pid == 0:
Bryan O'Sullivan
worker: fix a race in SIGINT handling...
r18708 signal.signal(signal.SIGINT, oldhandler)
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 try:
os.close(rfd)
for i, item in func(*(staticargs + (pargs,))):
os.write(wfd, '%d %s\n' % (i, item))
os._exit(0)
except KeyboardInterrupt:
os._exit(255)
Matt Mackall
worker: properly report errors from worker processes (issue3982)
r19408 # other exceptions are allowed to propagate, we rely
# on lock.py's pid checks to avoid release callbacks
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 pids.append(pid)
pids.reverse()
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 os.close(wfd)
fp = os.fdopen(rfd, 'rb', 0)
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 def killworkers():
# if one worker bails, there's no good reason to wait for the rest
for p in pids:
try:
os.kill(p, signal.SIGTERM)
Gregory Szorc
global: mass rewrite to use modern exception syntax...
r25660 except OSError as err:
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 if err.errno != errno.ESRCH:
raise
def waitforworkers():
Mads Kiilerich
cleanup: avoid _ for local unused tmp variables - that is reserved for i18n...
r22199 for _pid in pids:
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 st = _exitstatus(os.wait()[1])
Matt Mackall
worker: check problem state correctly (issue3982)...
r19406 if st and not problem[0]:
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 problem[0] = st
killworkers()
t = threading.Thread(target=waitforworkers)
t.start()
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 def cleanup():
signal.signal(signal.SIGINT, oldhandler)
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 t.join()
status = problem[0]
if status:
if status < 0:
os.kill(os.getpid(), -status)
sys.exit(status)
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 try:
for line in fp:
l = line.split(' ', 1)
yield int(l[0]), l[1][:-1]
except: # re-raises
Bryan O'Sullivan
worker: handle worker failures more aggressively...
r18709 killworkers()
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 cleanup()
raise
cleanup()
Bryan O'Sullivan
worker: on error, exit similarly to the first failing worker...
r18707 def _posixexitstatus(code):
'''convert a posix exit status into the same form returned by
os.spawnv
returns None if the process was stopped instead of exiting'''
if os.WIFEXITED(code):
return os.WEXITSTATUS(code)
elif os.WIFSIGNALED(code):
return -os.WTERMSIG(code)
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638 if os.name != 'nt':
_platformworker = _posixworker
Bryan O'Sullivan
worker: on error, exit similarly to the first failing worker...
r18707 _exitstatus = _posixexitstatus
Bryan O'Sullivan
worker: allow a function to be run in multiple worker processes...
r18638
Bryan O'Sullivan
worker: partition a list (of tasks) into equal-sized chunks
r18637 def partition(lst, nslices):
Gregory Szorc
worker: change partition strategy to every Nth element...
r28181 '''partition a list into N slices of roughly equal size
The current strategy takes every Nth element from the input. If
we ever write workers that need to preserve grouping in input
we should consider allowing callers to specify a partition strategy.
'''
for i in range(nslices):
yield lst[i::nslices]