def __init__(self, fullpath, vendor_flag=False):
cn.is_exist_path(fullpath)
self.fullpath = fullpath
self.vendor_flag = vendor_flag
self.basepath, self.dirpath = os.path.split(fullpath)
if cn.is_win():
self.statfile_info = cn.get_statfile(self.fullpath)
# 生成文件(FL_File对象)的列表
self.filelist = []
# 生成目录(FL_Dir对象)的列表
self.dirlist = []
for root, dirs, files in os.walk(self.fullpath, topdown=True):
for f in files:
if f in self.skip_list:
continue
fpath = os.path.join(root, f)
new_f = FL_File(fpath, self.fullpath, self.vendor_flag)
if cn.is_win():
new_f.set_info(self.statfile_info[new_f.rela_path])
self.filelist.append(new_f)
sys.stderr.write("Found file %-99s\r" % f)
for d in dirs:
dpath = os.path.join(root, d)
new_d = FL_Dir(dpath, self.fullpath, self.vendor_flag)
if cn.is_win():
new_d.set_info(self.statfile_info[new_d.rela_path])
self.dirlist.append(new_d)
cn.clean_line()
# 生成文件相对路径的列表
self.file_pathlist = [f.spath for f in self.filelist]
# 生成目录相对路径的列表
self.dir_pathlist = [f.spath for f in self.dirlist]
# 为文件和目录对象设置selabel属性
if not self.set_selabels(self.basepath):
if cn.is_win():
bootimg_ramdisk_path = cn.extract_bootimg(
os.path.join(self.basepath, "boot.img"))
if not self.set_selabels(bootimg_ramdisk_path):
raise Exception(
"Could not find (plat_)file_contexts(.bin)!"
" So we can not get selabel of files!")
else:
if os.system("ls -Z") == 0:
for f in (self.filelist + self.dirlist):
f.selabel = cn.get_selabel_linux(f.path)
else:
raise Exception("Can not get selabel with "
"\"ls -Z\" command!")
def watch_log_for(self, exprs, from_mark=None, timeout=600, process=None, verbose=False):
"""
Watch the log until one or more (regular) expression are found.
This methods when all the expressions have been found or the method
timeouts (a TimeoutError is then raised). On successful completion,
a list of pair (line matched, match object) is returned.
"""
elapsed = 0
tofind = [exprs] if isinstance(exprs, basestring) else exprs
tofind = [ re.compile(e) for e in tofind ]
matchings = []
reads = ""
if len(tofind) == 0:
return None
while not os.path.exists(self.logfilename()):
time.sleep(.5)
if process:
process.poll()
if process.returncode is not None:
self.print_process_output(self.name, process, verbose)
if process.returncode != 0:
raise RuntimeError() # Shouldn't reuse RuntimeError but I'm lazy
with open(self.logfilename()) as f:
if from_mark:
f.seek(from_mark)
while True:
# First, if we have a process to check, then check it.
# Skip on Windows - stdout/stderr is cassandra.bat
if not common.is_win():
if process:
process.poll()
if process.returncode is not None:
self.print_process_output(self.name, process, verbose)
if process.returncode != 0:
raise RuntimeError() # Shouldn't reuse RuntimeError but I'm lazy
line = f.readline()
if line:
reads = reads + line
for e in tofind:
m = e.search(line)
if m:
matchings.append((line, m))
tofind.remove(e)
if len(tofind) == 0:
return matchings[0] if isinstance(exprs, basestring) else matchings
else:
# yep, it's ugly
time.sleep(1)
elapsed = elapsed + 1
if elapsed > timeout:
raise TimeoutError(time.strftime("%d %b %Y %H:%M:%S", time.gmtime()) + " [" + self.name + "] Missing: " + str([e.pattern for e in tofind]) + ":\n" + reads)
if process:
process.poll()
if process.returncode is not None and process.returncode == 0:
return None
def check_win_requirements():
if common.is_win():
# Make sure ant.bat is in the path and executable before continuing
try:
process = subprocess.Popen('ant.bat', stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
except Exception as e:
sys.exit("ERROR! Could not find or execute ant.bat. Please fix this before attempting to run ccm on Windows.")
def __update_status(self):
if self.pid is None:
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
return
old_status = self.status
# os.kill on windows doesn't allow us to ping a process
if common.is_win():
self.__update_status_win()
else:
try:
os.kill(self.pid, 0)
except OSError as err:
if err.errno == errno.ESRCH:
# not running
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
elif err.errno == errno.EPERM:
# no permission to signal this process
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
else:
# some other error
raise err
else:
if self.status == Status.DOWN or self.status == Status.UNINITIALIZED:
self.status = Status.UP
if not old_status == self.status:
if old_status == Status.UP and self.status == Status.DOWN:
self.pid = None
self.__update_config()
def __update_log4j(self):
append_pattern = 'log4j.appender.R.File='
conf_file = os.path.join(self.get_conf_dir(), common.LOG4J_CONF)
log_file = os.path.join(self.get_path(), 'logs', 'system.log')
# log4j isn't partial to Windows \. I can't imagine why not.
if common.is_win():
log_file = re.sub("\\\\", "/", log_file)
common.replace_in_file(conf_file, append_pattern,
append_pattern + log_file)
# Setting the right log level
# Replace the global log level
if self.__global_log_level is not None:
append_pattern = 'log4j.rootLogger='
common.replace_in_file(
conf_file, append_pattern,
append_pattern + self.__global_log_level + ',stdout,R')
# Class specific log levels
for class_name in self.__classes_log_level:
logger_pattern = 'log4j.logger'
full_logger_pattern = logger_pattern + '.' + class_name + '='
common.replace_or_add_into_file_tail(
conf_file, full_logger_pattern,
full_logger_pattern + self.__classes_log_level[class_name])
def watch_log_for(self, exprs, from_mark=None, timeout=600, process=None, verbose=False):
"""
Watch the log until one or more (regular) expression are found.
This methods when all the expressions have been found or the method
timeouts (a TimeoutError is then raised). On successful completion,
a list of pair (line matched, match object) is returned.
"""
elapsed = 0
tofind = [exprs] if isinstance(exprs, string_types[0]) else exprs
tofind = [ re.compile(e) for e in tofind ]
matchings = []
reads = ""
if len(tofind) == 0:
return None
while not os.path.exists(self.logfilename()):
time.sleep(.5)
if process:
process.poll()
if process.returncode is not None:
self.print_process_output(self.name, process, verbose)
if process.returncode != 0:
raise RuntimeError() # Shouldn't reuse RuntimeError but I'm lazy
with open(self.logfilename()) as f:
if from_mark:
f.seek(from_mark)
while True:
# First, if we have a process to check, then check it.
# Skip on Windows - stdout/stderr is cassandra.bat
if not common.is_win():
if process:
process.poll()
if process.returncode is not None:
self.print_process_output(self.name, process, verbose)
if process.returncode != 0:
raise RuntimeError() # Shouldn't reuse RuntimeError but I'm lazy
line = f.readline()
if line:
reads = reads + line
for e in tofind:
m = e.search(line)
if m:
matchings.append((line, m))
tofind.remove(e)
if len(tofind) == 0:
return matchings[0] if isinstance(exprs, string_types[0]) else matchings
else:
# yep, it's ugly
time.sleep(1)
elapsed = elapsed + 1
if elapsed > timeout:
raise TimeoutError(time.strftime("%d %b %Y %H:%M:%S", time.gmtime()) + " [" + self.name + "] Missing: " + str([e.pattern for e in tofind]) + ":\n" + reads)
if process:
process.poll()
if process.returncode is not None and process.returncode == 0:
return None
def __init__(self, path, name, partitioner=None, cassandra_dir=None, create_directory=True, cassandra_version=None, verbose=False):
self.name = name
self.nodes = {}
self.seeds = []
self.partitioner = partitioner
self._config_options = {}
self.__log_level = "INFO"
self.__path = path
self.__version = None
if create_directory:
# we create the dir before potentially downloading to throw an error sooner if need be
os.mkdir(self.get_path())
try:
if cassandra_version is None:
# at this point, cassandra_dir should always not be None, but
# we keep this for backward compatibility (in loading old cluster)
if cassandra_dir is not None:
if common.is_win():
self.__cassandra_dir = cassandra_dir
else:
self.__cassandra_dir = os.path.abspath(cassandra_dir)
self.__version = self.__get_version_from_build()
else:
dir, v = repository.setup(cassandra_version, verbose)
self.__cassandra_dir = dir
self.__version = v if v is not None else self.__get_version_from_build()
if create_directory:
common.validate_cassandra_dir(self.__cassandra_dir)
self.__update_config()
except:
if create_directory:
shutil.rmtree(self.get_path())
raise
def __update_status(self):
if self.pid is None:
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
return
old_status = self.status
# os.kill on windows doesn't allow us to ping a process
if common.is_win():
self.__update_status_win()
else:
try:
os.kill(self.pid, 0)
except OSError, err:
if err.errno == errno.ESRCH:
# not running
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
elif err.errno == errno.EPERM:
# no permission to signal this process
if self.status == Status.UP or self.status == Status.DECOMMISIONNED:
self.status = Status.DOWN
else:
# some other error
raise err
else:
def unpack_img(self, img_path, is_new, is_vendor=False):
oon, sov = self.pars_init(is_new, is_vendor)
print("\nUnpacking %s Rom's %s.img..." % (oon, sov))
if cn.is_win():
spath = cn.extract_img(img_path)
cn.remove_path(img_path)
else:
spath = cn.mount_img(img_path)
return spath
def clean_temp():
print("\nCleaning temp files...")
for d in os.listdir(tempfile.gettempdir()):
if d.startswith("GOTAPGS_"):
if not cn.is_win():
for mdir in ("system_", "vendor_"):
os.system("sudo umount %s > /dev/null" %
os.path.join(tempfile.gettempdir(), d, mdir))
cn.remove_path(os.path.join(tempfile.gettempdir(), d))
def check_win_requirements():
if common.is_win():
# Make sure ant.bat is in the path and executable before continuing
try:
process = subprocess.Popen('ant.bat',
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT)
except Exception as e:
sys.exit(
"ERROR! Could not find or execute ant.bat. Please fix this before attempting to run ccm on Windows."
)
def clean_temp():
print("\nCleaning temp files...")
for d in os.listdir(tempfile.gettempdir()):
if d.startswith("GOTAPGS_"):
if not cn.is_win():
for mdir in ("system_", "vendor_"):
mount_path = os.path.join(tempfile.gettempdir(), d,
mdir)
os.system(
"mountpoint -q {0} && sudo umount {0}".format(
mount_path))
cn.remove_path(os.path.join(tempfile.gettempdir(), d))
def __init__(self, path, root_path):
self.uid = self.gid = self.perm = self.slink = self.sha1 = self.old_sha1 = self.selabel = ""
# 文件绝对路径
self.path = path
# 文件相对于"/"的路径
if root_path in self.path:
self.rela_path = self.path.replace(root_path, "",
1).replace("\\", "/")
else:
self.rela_path = ""
self.filename = os.path.split(path)[1]
if not common.is_win():
self.set_info(self.get_stat(self.path))
def stop(self, wait=True, wait_other_notice=False, gently=True):
"""
Stop the node.
- wait: if True (the default), wait for the Cassandra process to be
really dead. Otherwise return after having sent the kill signal.
- wait_other_notice: return only when the other live nodes of the
cluster have marked this node has dead.
- gently: Let Cassandra clean up and shut down properly. Otherwise do
a 'kill -9' which shuts down faster.
"""
if self.is_running():
if wait_other_notice:
#tstamp = time.time()
marks = [ (node, node.mark_log()) for node in self.cluster.nodes.values() if node.is_running() and node is not self ]
if common.is_win():
# Gentle on Windows is relative. WM_CLOSE is the best we get without external scripting
if gently:
os.system("taskkill /PID " + str(self.pid))
else:
os.system("taskkill /F /PID " + str(self.pid))
else:
if gently:
os.kill(self.pid, signal.SIGTERM)
else:
os.kill(self.pid, signal.SIGKILL)
if wait_other_notice:
for node, mark in marks:
node.watch_log_for_death(self, from_mark=mark)
#print node.name, "has marked", self.name, "down in " + str(time.time() - tstamp) + "s"
else:
time.sleep(.1)
still_running = self.is_running()
if still_running and wait:
wait_time_sec = 1
for i in xrange(0, 7):
# we'll double the wait time each try and cassandra should
# not take more than 1 minute to shutdown
time.sleep(wait_time_sec)
if not self.is_running():
return True
wait_time_sec = wait_time_sec * 2
raise NodeError("Problem stopping node %s" % self.name)
else:
return True
else:
return False
def stop(self, wait=True, wait_other_notice=False, gently=True):
"""
Stop the node.
- wait: if True (the default), wait for the Cassandra process to be
really dead. Otherwise return after having sent the kill signal.
- wait_other_notice: return only when the other live nodes of the
cluster have marked this node has dead.
- gently: Let Cassandra clean up and shut down properly. Otherwise do
a 'kill -9' which shuts down faster.
"""
if self.is_running():
if wait_other_notice:
#tstamp = time.time()
marks = [ (node, node.mark_log()) for node in list(self.cluster.nodes.values()) if node.is_running() and node is not self ]
if common.is_win():
# Gentle on Windows is relative. WM_CLOSE is the best we get without external scripting
if gently:
os.system("taskkill /PID " + str(self.pid))
else:
os.system("taskkill /F /PID " + str(self.pid))
else:
if gently:
os.kill(self.pid, signal.SIGTERM)
else:
os.kill(self.pid, signal.SIGKILL)
if wait_other_notice:
for node, mark in marks:
node.watch_log_for_death(self, from_mark=mark)
#print node.name, "has marked", self.name, "down in " + str(time.time() - tstamp) + "s"
else:
time.sleep(.1)
still_running = self.is_running()
if still_running and wait:
wait_time_sec = 1
for i in xrange(0, 7):
# we'll double the wait time each try and cassandra should
# not take more than 1 minute to shutdown
time.sleep(wait_time_sec)
if not self.is_running():
return True
wait_time_sec = wait_time_sec * 2
raise NodeError("Problem stopping node %s" % self.name)
else:
return True
else:
return False
def stress(self, stress_options):
stress = common.get_stress_bin(self.get_cassandra_dir())
livenodes = [ node.network_interfaces['storage'][0] for node in list(self.nodes.values()) if node.is_live() ]
if len(livenodes) == 0:
print_("No live node")
return
args = [ stress, '-d', ",".join(livenodes) ] + stress_options
try:
# need to set working directory for env on Windows
if common.is_win():
subprocess.call(args, cwd=common.parse_path(stress))
else:
subprocess.call(args)
except KeyboardInterrupt:
pass
return self
def validate_cassandra_dir(cassandra_dir):
if cassandra_dir is None:
raise ArgumentError('Undefined cassandra directory')
# Windows requires absolute pathing on cassandra dir - abort if specified cygwin style
if common.is_win():
if ':' not in cassandra_dir:
raise ArgumentError('%s does not appear to be a cassandra source directory. Please use absolute pathing (e.g. C:/cassandra.' % cassandra_dir)
bin_dir = os.path.join(cassandra_dir, CASSANDRA_BIN_DIR)
conf_dir = os.path.join(cassandra_dir, CASSANDRA_CONF_DIR)
cnd = os.path.exists(bin_dir)
cnd = cnd and os.path.exists(conf_dir)
cnd = cnd and os.path.exists(os.path.join(conf_dir, CASSANDRA_CONF))
if not cnd:
raise ArgumentError('%s does not appear to be a cassandra source directory' % cassandra_dir)
def is_metric_valid(metric):
if 'name' not in metric or 'cmd' not in metric:
logging.warn('incompleted metric definition %s', metric)
return False
name = metric['name']
os = metric.get('os')
cmd = metric['cmd']
checkcmd = 'which'
if is_win():
checkcmd = 'where'
if is_sunos():
checkcmd = 'type'
if os is None or os == ostype():
valid = call([checkcmd, cmd[0]]) == 0
else:
valid = False
logging.info('check metric %s with os=%s -> %s', name, os, valid)
return valid
def __init__(self,
path,
name,
partitioner=None,
cassandra_dir=None,
create_directory=True,
cassandra_version=None,
verbose=False):
self.name = name
self.nodes = {}
self.seeds = []
self.partitioner = partitioner
self._config_options = {}
self.__log_level = "INFO"
self.__path = path
self.__version = None
if create_directory:
# we create the dir before potentially downloading to throw an error sooner if need be
os.mkdir(self.get_path())
try:
if cassandra_version is None:
# at this point, cassandra_dir should always not be None, but
# we keep this for backward compatibility (in loading old cluster)
if cassandra_dir is not None:
if common.is_win():
self.__cassandra_dir = cassandra_dir
else:
self.__cassandra_dir = os.path.abspath(cassandra_dir)
self.__version = self.__get_version_from_build()
else:
dir, v = repository.setup(cassandra_version, verbose)
self.__cassandra_dir = dir
self.__version = v if v is not None else self.__get_version_from_build(
)
if create_directory:
common.validate_cassandra_dir(self.__cassandra_dir)
self.__update_config()
except:
if create_directory:
shutil.rmtree(self.get_path())
raise
def stress(self, stress_options):
stress = common.get_stress_bin(self.get_cassandra_dir())
livenodes = [
node.network_interfaces['storage'][0]
for node in self.nodes.values() if node.is_live()
]
if len(livenodes) == 0:
print "No live node"
return
args = [stress, '-d', ",".join(livenodes)] + stress_options
try:
# need to set working directory for env on Windows
if common.is_win():
subprocess.call(args, cwd=common.parse_path(stress))
else:
subprocess.call(args)
except KeyboardInterrupt:
pass
return self
def validate_cassandra_dir(cassandra_dir):
if cassandra_dir is None:
raise ArgumentError('Undefined cassandra directory')
# Windows requires absolute pathing on cassandra dir - abort if specified cygwin style
if common.is_win():
if ':' not in cassandra_dir:
raise ArgumentError(
'%s does not appear to be a cassandra source directory. Please use absolute pathing (e.g. C:/cassandra.'
% cassandra_dir)
bin_dir = os.path.join(cassandra_dir, CASSANDRA_BIN_DIR)
conf_dir = os.path.join(cassandra_dir, CASSANDRA_CONF_DIR)
cnd = os.path.exists(bin_dir)
cnd = cnd and os.path.exists(conf_dir)
cnd = cnd and os.path.exists(os.path.join(conf_dir, CASSANDRA_CONF))
if not cnd:
raise ArgumentError(
'%s does not appear to be a cassandra source directory' %
cassandra_dir)
def __update_log4j(self):
append_pattern='log4j.appender.R.File='
conf_file = os.path.join(self.get_conf_dir(), common.LOG4J_CONF)
log_file = os.path.join(self.get_path(), 'logs', 'system.log')
# log4j isn't partial to Windows \. I can't imagine why not.
if common.is_win():
log_file = re.sub("\\\\", "/", log_file)
common.replace_in_file(conf_file, append_pattern, append_pattern + log_file)
# Setting the right log level
# Replace the global log level
if self.__global_log_level is not None:
append_pattern='log4j.rootLogger='
common.replace_in_file(conf_file, append_pattern, append_pattern + self.__global_log_level + ',stdout,R')
# Class specific log levels
for class_name in self.__classes_log_level:
logger_pattern='log4j.logger'
full_logger_pattern = logger_pattern + '.' + class_name + '='
common.replace_or_add_into_file_tail(conf_file, full_logger_pattern, full_logger_pattern + self.__classes_log_level[class_name])
def __init__(self, path, root_path, vendor_flag):
self.uid = self.gid = self.perm = self.slink = ""
# 文件绝对路径
self.path = path
# 文件"根"路径
self.root_path = root_path
self.vendor_flag = vendor_flag
# 文件父目录路径 & 文件名
self.pare_path, self.name = os.path.split(path)
# 文件相对"根"路径的相对路径
if root_path in self.path:
self.rela_path = self.path.replace(root_path, "", 1)[1:]
else:
self.rela_path = ""
# 文件在卡刷包中存在的路径
if self.vendor_flag:
self.spath = "/vendor/" + self.rela_path.replace("\\", "/")
else:
self.spath = "/system/" + self.rela_path.replace("\\", "/")
if not cn.is_win():
self.set_info(self.get_stat(self.path))
def start(self,
join_ring=True,
no_wait=False,
verbose=False,
update_pid=True,
wait_other_notice=False,
replace_token=None,
replace_address=None,
jvm_args=[],
wait_for_binary_proto=False,
profile_options=None,
use_jna=False):
"""
Start the node. Options includes:
- join_ring: if false, start the node with -Dcassandra.join_ring=False
- no_wait: by default, this method returns when the node is started and listening to clients.
If no_wait=True, the method returns sooner.
- wait_other_notice: if True, this method returns only when all other live node of the cluster
have marked this node UP.
- replace_token: start the node with the -Dcassandra.replace_token option.
- replace_address: start the node with the -Dcassandra.replace_address option.
"""
if self.is_running():
raise NodeError("%s is already running" % self.name)
for itf in self.network_interfaces.values():
if itf is not None and replace_address is None:
common.check_socket_available(itf)
if wait_other_notice:
marks = [(node, node.mark_log())
for node in self.cluster.nodes.values()
if node.is_running()]
cdir = self.get_cassandra_dir()
cass_bin = common.join_bin(cdir, 'bin', 'cassandra')
# Copy back the cassandra scripts since profiling may have modified it the previous time
shutil.copy(cass_bin, self.get_bin_dir())
cass_bin = common.join_bin(self.get_path(), 'bin', 'cassandra')
# If Windows, change entries in .bat file to split conf from binaries
if common.is_win():
self.__clean_bat()
if profile_options is not None:
config = common.get_config()
if not 'yourkit_agent' in config:
raise NodeError(
"Cannot enable profile. You need to set 'yourkit_agent' to the path of your agent in a ~/.ccm/config"
)
cmd = '-agentpath:%s' % config['yourkit_agent']
if 'options' in profile_options:
cmd = cmd + '=' + profile_options['options']
print cmd
# Yes, it's fragile as shit
pattern = r'cassandra_parms="-Dlog4j.configuration=log4j-server.properties -Dlog4j.defaultInitOverride=true'
common.replace_in_file(cass_bin, pattern,
' ' + pattern + ' ' + cmd + '"')
os.chmod(cass_bin, os.stat(cass_bin).st_mode | stat.S_IEXEC)
env = common.make_cassandra_env(cdir, self.get_path())
pidfile = os.path.join(self.get_path(), 'cassandra.pid')
args = [
cass_bin, '-p', pidfile,
'-Dcassandra.join_ring=%s' % str(join_ring)
]
if replace_token is not None:
args.append('-Dcassandra.replace_token=%s' % str(replace_token))
if replace_address is not None:
args.append('-Dcassandra.replace_address=%s' %
str(replace_address))
if use_jna is False:
args.append('-Dcassandra.boot_without_jna=true')
args = args + jvm_args
process = None
if common.is_win():
# clean up any old dirty_pid files from prior runs
if (os.path.isfile(self.get_path() + "/dirty_pid.tmp")):
os.remove(self.get_path() + "/dirty_pid.tmp")
process = subprocess.Popen(args,
cwd=self.get_bin_dir(),
env=env,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
else:
process = subprocess.Popen(args,
env=env,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
# Our modified batch file writes a dirty output with more than just the pid - clean it to get in parity
# with *nix operation here.
if common.is_win():
self.__clean_win_pid()
self._update_pid(process)
elif update_pid:
if no_wait:
time.sleep(
2
) # waiting 2 seconds nevertheless to check for early errors and for the pid to be set
else:
for line in process.stdout:
if verbose:
print line.rstrip('\n')
self._update_pid(process)
if not self.is_running():
raise NodeError("Error starting node %s" % self.name, process)
if wait_other_notice:
for node, mark in marks:
node.watch_log_for_alive(self, from_mark=mark)
if wait_for_binary_proto:
self.watch_log_for("Starting listening for CQL clients")
# we're probably fine at that point but just wait some tiny bit more because
# the msg is logged just before starting the binary protocol server
time.sleep(0.2)
return process
def start(self,
join_ring=True,
no_wait=False,
verbose=False,
update_pid=True,
wait_other_notice=False,
replace_token=None,
replace_address=None,
jvm_args=[],
wait_for_binary_proto=False,
profile_options=None,
use_jna=False):
"""
Start the node. Options includes:
- join_ring: if false, start the node with -Dcassandra.join_ring=False
- no_wait: by default, this method returns when the node is started and listening to clients.
If no_wait=True, the method returns sooner.
- wait_other_notice: if True, this method returns only when all other live node of the cluster
have marked this node UP.
- replace_token: start the node with the -Dcassandra.replace_token option.
- replace_address: start the node with the -Dcassandra.replace_address option.
"""
if self.is_running():
raise NodeError("%s is already running" % self.name)
for itf in list(self.network_interfaces.values()):
if itf is not None and replace_address is None:
common.check_socket_available(itf)
if wait_other_notice:
marks = [ (node, node.mark_log()) for node in list(self.cluster.nodes.values()) if node.is_running() ]
cdir = self.get_cassandra_dir()
cass_bin = common.join_bin(cdir, 'bin', 'cassandra')
# Copy back the cassandra scripts since profiling may have modified it the previous time
shutil.copy(cass_bin, self.get_bin_dir())
cass_bin = common.join_bin(self.get_path(), 'bin', 'cassandra')
# If Windows, change entries in .bat file to split conf from binaries
if common.is_win():
self.__clean_bat()
if profile_options is not None:
config = common.get_config()
if not 'yourkit_agent' in config:
raise NodeError("Cannot enable profile. You need to set 'yourkit_agent' to the path of your agent in a ~/.ccm/config")
cmd = '-agentpath:%s' % config['yourkit_agent']
if 'options' in profile_options:
cmd = cmd + '=' + profile_options['options']
print_(cmd)
# Yes, it's fragile as shit
pattern=r'cassandra_parms="-Dlog4j.configuration=log4j-server.properties -Dlog4j.defaultInitOverride=true'
common.replace_in_file(cass_bin, pattern, ' ' + pattern + ' ' + cmd + '"')
os.chmod(cass_bin, os.stat(cass_bin).st_mode | stat.S_IEXEC)
env = common.make_cassandra_env(cdir, self.get_path())
pidfile = os.path.join(self.get_path(), 'cassandra.pid')
args = [ cass_bin, '-p', pidfile, '-Dcassandra.join_ring=%s' % str(join_ring) ]
if replace_token is not None:
args.append('-Dcassandra.replace_token=%s' % str(replace_token))
if replace_address is not None:
args.append('-Dcassandra.replace_address=%s' % str(replace_address))
if use_jna is False:
args.append('-Dcassandra.boot_without_jna=true')
args = args + jvm_args
process = None
if common.is_win():
process = subprocess.Popen(args, cwd=self.get_bin_dir(), env=env, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
else:
process = subprocess.Popen(args, env=env, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
# Our modified batch file writes a dirty output with more than just the pid - clean it to get in parity
# with *nix operation here.
if common.is_win():
# short delay could give us a false positive on a node being started as it could die and delete the pid file
# after we check, however dtests have assumptions on how long the starting process takes.
time.sleep(.5)
self.__clean_win_pid()
self._update_pid(process)
elif update_pid:
if no_wait:
time.sleep(2) # waiting 2 seconds nevertheless to check for early errors and for the pid to be set
else:
for line in process.stdout:
if verbose:
print_(line.rstrip('\n'))
self._update_pid(process)
if not self.is_running():
raise NodeError("Error starting node %s" % self.name, process)
if wait_other_notice:
for node, mark in marks:
node.watch_log_for_alive(self, from_mark=mark)
if wait_for_binary_proto:
self.watch_log_for("Starting listening for CQL clients")
# we're probably fine at that point but just wait some tiny bit more because
# the msg is logged just before starting the binary protocol server
time.sleep(0.2)
return process