我正在编写一些监控工具，一个这样的工具处理NFS挂载的文件系统列表，并尝试将测试文件写入每个共享。我为每个被测试的文件管理器生成一个线程。在一定的超时后，我想终止任何仍在运行的线程。我不需要从线程中收集输出，我不需要加入或连接到任何线程，我只需要停止在超时后仍然运行的任何线程。
所以我故意在每个线程中运行一个sleep（300），所以我知道没有问题，当我遍历所有活动线程并尝试杀死它们时，每个线程中产生的进程仍然在最后运行。这适用于5-20个线程的任何地方，但最终在一个线程上失败，说self.pid不再有效。发生这种情况的线程是随机的，不一定是与上一次运行相同的线程。
class NFSWriteTestThread(threading.Thread):
def __init__(self, filer, base_mnt_point, number):
super(NFSWriteTestThread, self).__init__()
self.tname = filer
self.tnum = number
self.filer = filer
self.mntpt = base_mnt_point
self.process = None
self.pid = None
def run(self):
start = time.time()
# self.process = subprocess.Popen(['/bin/dd', 'if=/dev/zero', 'bs=1M', 'count=5', 'of=' + self.testfile], shell=False)
self.process = subprocess.Popen(['/bin/sleep', '300'], shell=False)
time.sleep(1)
logger.debug("DEBUG: %s=%d" % (self.tname, self.process.pid))
self.pid = self.process.pid
logger.info(" NFS write test command initiaited on '%s', pid=%d" % (self.filer, self.pid))
self.process.wait()
# self.output, self.error = self.process.communicate()
end = time.time()
logger.info(" NFS write test for '%s' completed in %d seconds" % (self.filer, end - start))
return
def getThreadName(self):
return self.tname
def getThreadNum(self):
return self.tnum
def getThreadPID(self):
if self.pid:
return self.pid
else:
return "unknown"
def isAlive(self):
if not self.process:
logger.debug("Error: self.process is invalid (%s)" % type(self.process))
# if self.process.poll():
# logger.info("NFS write test operation for thread '%s' is still active" % self.filer)
# else:
# logger.info("NFS write test operation for thread '%s' is inactive" % self.filer)
return
def terminate(self):
os.kill(self.process.pid, signal.SIGTERM)
return
def kill(self):
os.kill(self.process.pid, signal.SIGKILL)
return
def initLogging(config):
logfile = os.path.join(config['logdir'], config['logfilename'])
fformat = logging.Formatter('%(asctime)s %(message)s', "%Y-%m-%d %H:%M:%S %Z")
cformat = logging.Formatter('%(asctime)s %(message)s', "%Y-%m-%d %H:%M:%S %Z")
clogger = None
flogger = None
if config['debug']:
loglevel = logging.DEBUG
if not os.path.exists(config['logdir']):
os.makedirs(config['logdir'])
os.chmod(config['logdir'], 0700)
os.chown(config['logdir'], 0, 0)
try:
logger = logging.getLogger('main')
logger.setLevel(logging.DEBUG)
# Define a file logger
flogger = logging.FileHandler(logfile, 'w')
flogger.setLevel(logging.DEBUG)
flogger.setFormatter(fformat)
logger.addHandler(flogger)
# Define a console logger if verbose
if config['verbose']:
clogger = logging.StreamHandler()
clogger.setLevel(logging.DEBUG)
clogger.setFormatter(cformat)
logger.addHandler(clogger)
except Exception, error:
print "Error: Unable to initialize file logging: %s" % error
sys.exit(1)
logger.info("Script initiated.")
logger.info("Using the following configuration:")
for key, value in sorted(config.iteritems()):
logger.info(" %20s = '%-s'" % (key, value))
return logger
def parseConfigFile(cfg):
if not os.path.isfile(cfg['cfgfile']) or not os.access(cfg['cfgfile'], os.R_OK):
print "Error: '%s' does not exist or is not readable, terminating." % cfg['cfgfile']
sys.exit(1)
config = SafeConfigParser()
config.read(cfg['cfgfile'])
_cfg = dict(config.items(cfg['cfgfilestanza']))
_cfgfilers = config.get(cfg['cfgfilestanza'], 'managed_filers')
_tmpfilers = _cfgfilers.split(',')
# populate a list containing all filers which will be meged into the global cfg[] dict
_cfg['filers'] = []
for _f in _tmpfilers:
_cfg['filers'].append(_f.strip())
return _cfg
logger = initLogging(cfg)
cfg.update(parseConfigFile(cfg))
threads = []
numThreads = 0
for filer in cfg['filers']:
numThreads += 1
logger.debug(" spawning NFS wite test thread for '%s', thread number %s" % (filer, numThreads))
t = NFSWriteTestThread(filer, cfg['base_mnt_point'], numThreads)
t.start()
threads.append(t)
# time.sleep(1)
logger.info("spawned %d NFS write test child threads" % numThreads)
logger.info("sleeping for %d seconds" % cfg['timeout'])
time.sleep(cfg['timeout'])
if (threading.activeCount() > 1):
logger.info("there are %d NFS write test threads active after the timeout:" % (threading.activeCount() - 1))
for thr in threading.enumerate():
logger.debug("theadname=%s" % thr.name)
if re.match("MainThread", thr.getName()):
pass
else:
logger.info("thread '%s' (thread %d) is still alive" % (thr.getThreadName(), thr.getThreadNum()))
# thr.isAlive()
logger.info("killing thread for '%s' (pid=XX) with SIGTERM" % (thr.getThreadName()))
# logger.info("killing thread for '%s' (pid=%d) with SIGTERM" % (thr.getThreadName(), thr.getThreadPID()))
thr.kill()
logger.info("Script complete")
sys.exit(0)
在这里你可以看到输出：
2014-11-10 09:00:22 CST there are 173 NFS write test threads active after the timeout:
2014-11-10 09:00:22 CST theadname=Thread-165
2014-11-10 09:00:22 CST thread 'hostname1' (thread 165) is still alive
2014-11-10 09:00:22 CST killing thread for 'hostname1' (pid=XX) with SIGTERM
2014-11-10 09:00:22 CST theadname=Thread-97
2014-11-10 09:00:22 CST thread 'hostname2' (thread 97) is still alive
2014-11-10 09:00:22 CST NFS write test for 'hostname1' completed in 60 seconds
2014-11-10 09:00:22 CST killing thread for 'hostname2' (pid=XX) with SIGTERM
2014-11-10 09:00:22 CST theadname=Thread-66
2014-11-10 09:00:22 CST thread 'hostname3' (thread 66) is still alive
2014-11-10 09:00:22 CST NFS write test for 'hostname2' completed in 60 seconds
2014-11-10 09:00:22 CST killing thread for 'hostname3' (pid=XX) with SIGTERM
2014-11-10 09:00:22 CST theadname=Thread-121
2014-11-10 09:00:22 CST thread 'hostname4' (thread 121) is still alive
2014-11-10 09:00:22 CST killing thread for 'hostname4' (pid=XX) with SIGTERM
Traceback (most recent call last):
2014-11-10 09:00:22 CST NFS write test for 'hostname3' completed in 60 seconds
File "./NFSWriteTestCheck.py", line 199, in <module>
thr.kill()
File "./NFSWriteTestCheck.py", line 84, in kill
os.kill(self.process.pid, signal.SIGKILL)
AttributeError: 'NoneType' object has no attribute
在显示此错误时，进程仍在运行，在shell中使用ps进行验证。为什么线程对象不再有效？在抛出此错误时，线程执行应该在此时：
self.process.wait（）
只是在这里挠头，想知道我是不是在碰到一个小虫或什么的。
答案 0 :(得分：1)
你不能停止线程，你只能打断他们正在做的事情。在您的情况下，线程正在等待subprocess.call。设置事件无效，因为您的线程不会等待事件。这里的解决方案是杀死子进程，这意味着你需要Popen对象。
我将实现直接放入run方法中，以便Popen对象很方便。
class NFSWriteTestThread(threading.Thread):
def __init__(self, filer, mntpoint, number):
super(NFSWriteTestThread, self).__init__()
self.name = filer
self.filer = filer
self.mntpt = mntpoint
self.tnum = number
self._proc = None
def run(self):
testfile = "%s/%s/test/test.%s" % (mountpoint, filer, filer)
testcmd = "/bin/bash -c '/bin/dd if=/dev/zero bs=1024 count=1024 of=" + testfile + " >/dev/null 2>/dev/null; sleep 120'"
self._proc = subprocess.Popen(testcmd, shell=True)
self._proc.wait()
return
def getName(self):
return self.name
def kill(self):
if self._proc:
self._proc.terminate()
def stopped(self):
if self._proc:
return self._proc.poll() is not None
else:
return True