def run(self):
"""
Run the promise
This will first load the promise module (which will update process sys.path)
"""
try:
os.chdir(self.partition_folder)
promise_started = False
if self.uid and self.gid:
dropPrivileges(self.uid, self.gid, logger=self.logger)
mkdir_p(self.cache_folder)
if self.wrap_promise:
promise_instance = WrapPromise(self.argument_dict)
else:
self._createInitFile()
promise_module = self._loadPromiseModule()
promise_instance = promise_module.RunPromise(self.argument_dict)
if not hasattr(promise_instance, 'isAnomalyDetected') or not \
hasattr(promise_instance, 'isTested') or \
(promise_instance.isAnomalyDetected() and self.check_anomaly) or \
(promise_instance.isTested() and not self.check_anomaly):
# if the promise will run, we save execution timestamp
promise_started = True
self.updatePromiseCache(
WrapPromise if self.wrap_promise else promise_module.RunPromise,
promise_instance,
started=promise_started)
promise_instance.run(self.check_anomaly, self.allow_bang)
except Exception:
self.logger.error(traceback.format_exc())
raise
def _checkPromises(self, computer_partition):
self.logger.info("Checking promises...")
instance_path = os.path.join(self.instance_root, computer_partition.getId())
uid, gid = None, None
stat_info = os.stat(instance_path)
#stat sys call to get statistics informations
uid = stat_info.st_uid
gid = stat_info.st_gid
promise_present = False
# Get the list of promises
promise_dir = os.path.join(instance_path, 'etc', 'promise')
if os.path.exists(promise_dir) and os.path.isdir(promise_dir):
# Check whether every promise is kept
for promise in os.listdir(promise_dir):
promise_present = True
command = [os.path.join(promise_dir, promise)]
promise = os.path.basename(command[0])
self.logger.info("Checking promise '%s'.", promise)
process_handler = subprocess.Popen(command,
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=instance_path,
env=None if sys.platform == 'cygwin' else {},
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stdin=subprocess.PIPE)
process_handler.stdin.flush()
process_handler.stdin.close()
process_handler.stdin = None
# Check if the promise finished every tenth of second,
# but timeout after promise_timeout.
sleep_time = 0.1
increment_limit = int(self.promise_timeout / sleep_time)
for current_increment in range(0, increment_limit):
if process_handler.poll() is None:
time.sleep(sleep_time)
continue
if process_handler.poll() == 0:
# Success!
break
else:
stderr = process_handler.communicate()[1]
if stderr is None:
stderr = "No error output from '%s'." % promise
else:
stderr = "Promise '%s':" % promise + stderr
raise Slapgrid.PromiseError(stderr)
else:
process_handler.terminate()
raise Slapgrid.PromiseError("The promise '%s' timed out" % promise)
if not promise_present:
self.logger.info("No promise.")
def _checkPromises(self, computer_partition):
self.logger.info("Checking promises...")
instance_path = os.path.join(self.instance_root, computer_partition.getId())
uid, gid = None, None
stat_info = os.stat(instance_path)
#stat sys call to get statistics informations
uid = stat_info.st_uid
gid = stat_info.st_gid
promise_present = False
# Get the list of promises
promise_dir = os.path.join(instance_path, 'etc', 'promise')
if os.path.exists(promise_dir) and os.path.isdir(promise_dir):
# Check whether every promise is kept
for promise in os.listdir(promise_dir):
promise_present = True
command = [os.path.join(promise_dir, promise)]
promise = os.path.basename(command[0])
self.logger.info("Checking promise %r.", promise)
process_handler = subprocess.Popen(command,
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=instance_path,
env=None if sys.platform == 'cygwin' else {},
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
stdin=subprocess.PIPE)
process_handler.stdin.flush()
process_handler.stdin.close()
process_handler.stdin = None
time.sleep(self.promise_timeout)
if process_handler.poll() is None:
process_handler.terminate()
raise Slapgrid.PromiseError("The promise %r timed out" % promise)
elif process_handler.poll() != 0:
stderr = process_handler.communicate()[1]
if stderr is None:
stderr = 'No error output from %r.' % promise
else:
stderr = 'Promise %r:' % promise + stderr
raise Slapgrid.PromiseError(stderr)
if not promise_present:
self.logger.info("No promise.")
def destroy(self):
"""Destroys the partition and makes it available for subsequent use."
"""
self.logger.info("Destroying Computer Partition %s..."
% self.computer_partition.getId())
# Launches "destroy" binary if exists
destroy_executable_location = os.path.join(self.instance_path, 'sbin',
'destroy')
if os.path.exists(destroy_executable_location):
uid, gid = self.getUserGroupId()
self.logger.debug('Invoking %r' % destroy_executable_location)
process_handler = SlapPopen([destroy_executable_location],
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=self.instance_path,
env=getCleanEnvironment(logger=self.logger,
home_path=pwd.getpwuid(uid).pw_dir),
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
logger=self.logger)
if process_handler.returncode is None or process_handler.returncode != 0:
message = 'Failed to destroy Computer Partition in %r.' % \
self.instance_path
self.logger.error(message)
raise subprocess.CalledProcessError(message, process_handler.output)
# Manually cleans what remains
try:
for f in [self.key_file, self.cert_file]:
if f:
if os.path.exists(f):
os.unlink(f)
# better to manually remove symlinks because rmtree might choke on them
sr_symlink = os.path.join(self.instance_path, 'software_release')
if os.path.islink(sr_symlink):
os.unlink(sr_symlink)
for root, dirs, file_list in os.walk(self.instance_path):
for directory in dirs:
shutil.rmtree(os.path.join(self.instance_path, directory))
for file in file_list:
os.remove(os.path.join(self.instance_path, file))
if os.path.exists(self.supervisord_partition_configuration_path):
os.remove(self.supervisord_partition_configuration_path)
self.updateSupervisor()
except IOError as exc:
raise IOError("I/O error while freeing partition (%s): %s" % (self.instance_path, exc))
def agregateAndSendUsage(self):
"""Will agregate usage from each Computer Partition.
"""
# Prepares environment
self.checkEnvironmentAndCreateStructure()
self._launchSupervisord()
slap_computer_usage = self.slap.registerComputer(self.computer_id)
computer_partition_usage_list = []
self.logger.info('Aggregating and sending usage reports...')
#We retrieve XSD models
try:
computer_consumption_model = \
pkg_resources.resource_string(
'slapos.slap',
'doc/computer_consumption.xsd')
except IOError:
computer_consumption_model = \
pkg_resources.resource_string(
__name__,
'../../../../slapos/slap/doc/computer_consumption.xsd')
try:
partition_consumption_model = \
pkg_resources.resource_string(
'slapos.slap',
'doc/partition_consumption.xsd')
except IOError:
partition_consumption_model = \
pkg_resources.resource_string(
__name__,
'../../../../slapos/slap/doc/partition_consumption.xsd')
clean_run = True
# Loop over the different computer partitions
computer_partition_list = self.FilterComputerPartitionList(
slap_computer_usage.getComputerPartitionList())
for computer_partition in computer_partition_list:
try:
computer_partition_id = computer_partition.getId()
#We want to execute all the script in the report folder
instance_path = os.path.join(self.instance_root,
computer_partition.getId())
report_path = os.path.join(instance_path, 'etc', 'report')
if os.path.isdir(report_path):
script_list_to_run = os.listdir(report_path)
else:
script_list_to_run = []
#We now generate the pseudorandom name for the xml file
# and we add it in the invocation_list
f = tempfile.NamedTemporaryFile()
name_xml = '%s.%s' % ('slapreport', os.path.basename(f.name))
path_to_slapreport = os.path.join(instance_path, 'var', 'xml_report',
name_xml)
failed_script_list = []
for script in script_list_to_run:
invocation_list = []
invocation_list.append(os.path.join(instance_path, 'etc', 'report',
script))
#We add the xml_file name to the invocation_list
#f = tempfile.NamedTemporaryFile()
#name_xml = '%s.%s' % ('slapreport', os.path.basename(f.name))
#path_to_slapreport = os.path.join(instance_path, 'var', name_xml)
invocation_list.append(path_to_slapreport)
#Dropping privileges
uid, gid = None, None
stat_info = os.stat(instance_path)
#stat sys call to get statistics informations
uid = stat_info.st_uid
gid = stat_info.st_gid
process_handler = SlapPopen(invocation_list,
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=os.path.join(instance_path, 'etc', 'report'),
env=None,
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
logger=self.logger)
if process_handler.returncode is None:
process_handler.kill()
if process_handler.returncode != 0:
clean_run = False
failed_script_list.append("Script %r failed." % script)
self.logger.warning('Failed to run %r' % invocation_list)
if len(failed_script_list):
computer_partition.error('\n'.join(failed_script_list), logger=self.logger)
# Whatever happens, don't stop processing other instances
except Exception:
self.logger.exception('Cannot run usage script(s) for %r:' %
computer_partition.getId())
#Now we loop through the different computer partitions to report
report_usage_issue_cp_list = []
for computer_partition in computer_partition_list:
try:
filename_delete_list = []
computer_partition_id = computer_partition.getId()
instance_path = os.path.join(self.instance_root, computer_partition_id)
dir_reports = os.path.join(instance_path, 'var', 'xml_report')
#The directory xml_report contain a number of files equal
#to the number of software instance running inside the same partition
if os.path.isdir(dir_reports):
filename_list = os.listdir(dir_reports)
else:
filename_list = []
#self.logger.debug('name List %s' % filename_list)
for filename in filename_list:
file_path = os.path.join(dir_reports, filename)
if os.path.exists(file_path):
usage = open(file_path, 'r').read()
#We check the validity of xml content of each reports
if not self.validateXML(usage, partition_consumption_model):
self.logger.info('WARNING: The XML file %s generated by slapreport is '
'not valid - This report is left as is at %s where you can '
'inspect what went wrong ' % (filename, dir_reports))
# Warn the SlapOS Master that a partition generates corrupted xml
# report
else:
computer_partition_usage = self.slap.registerComputerPartition(
self.computer_id, computer_partition_id)
computer_partition_usage.setUsage(usage)
computer_partition_usage_list.append(computer_partition_usage)
filename_delete_list.append(filename)
else:
self.logger.debug('Usage report %r not found, ignored' % file_path)
#After sending the aggregated file we remove all the valid xml reports
for filename in filename_delete_list:
os.remove(os.path.join(dir_reports, filename))
# Whatever happens, don't stop processing other instances
except Exception:
self.logger.exception('Cannot run usage script(s) for %r:' %
computer_partition.getId())
for computer_partition_usage in computer_partition_usage_list:
self.logger.info('computer_partition_usage_list: %s - %s' %
(computer_partition_usage.usage, computer_partition_usage.getId()))
#If there is, at least, one report
if computer_partition_usage_list != []:
try:
#We generate the final XML report with asXML method
computer_consumption = self.asXML(computer_partition_usage_list)
self.logger.info('Final xml report: %s' % computer_consumption)
#We test the XML report before sending it
if self.validateXML(computer_consumption, computer_consumption_model):
self.logger.info('XML file generated by asXML is valid')
slap_computer_usage.reportUsage(computer_consumption)
else:
self.logger.info('XML file generated by asXML is not valid !')
raise ValueError('XML file generated by asXML is not valid !')
except Exception:
issue = "Cannot report usage for %r: %s" % (
computer_partition.getId(),
traceback.format_exc())
self.logger.info(issue)
computer_partition.error(issue, logger=self.logger)
report_usage_issue_cp_list.append(computer_partition_id)
for computer_partition in computer_partition_list:
if computer_partition.getState() == COMPUTER_PARTITION_DESTROYED_STATE:
try:
computer_partition_id = computer_partition.getId()
try:
software_url = computer_partition.getSoftwareRelease().getURI()
software_path = os.path.join(self.software_root, md5digest(software_url))
except (NotFoundError, TypeError):
software_url = None
software_path = None
local_partition = Partition(
software_path=software_path,
instance_path=os.path.join(self.instance_root,
computer_partition.getId()),
supervisord_partition_configuration_path=os.path.join(
self.supervisord_configuration_directory, '%s.conf' %
computer_partition_id),
supervisord_socket=self.supervisord_socket,
computer_partition=computer_partition,
computer_id=self.computer_id,
partition_id=computer_partition_id,
server_url=self.master_url,
software_release_url=software_url,
certificate_repository_path=self.certificate_repository_path,
buildout=self.buildout,
logger=self.logger)
local_partition.stop()
try:
computer_partition.stopped()
except (SystemExit, KeyboardInterrupt):
computer_partition.error(traceback.format_exc(), logger=self.logger)
raise
except Exception:
pass
if computer_partition.getId() in report_usage_issue_cp_list:
self.logger.info('Ignoring destruction of %r, as no report usage was sent' %
computer_partition.getId())
continue
local_partition.destroy()
except (SystemExit, KeyboardInterrupt):
computer_partition.error(traceback.format_exc(), logger=self.logger)
raise
except Exception:
clean_run = False
self.logger.exception('')
exc = traceback.format_exc()
computer_partition.error(exc, logger=self.logger)
try:
computer_partition.destroyed()
except NotFoundError:
self.logger.debug('Ignored slap error while trying to inform about '
'destroying not fully configured Computer Partition %r' %
computer_partition.getId())
except ServerError as server_error:
self.logger.debug('Ignored server error while trying to inform about '
'destroying Computer Partition %r. Error is:\n%r' %
(computer_partition.getId(), server_error.args[0]))
self.logger.info('Finished usage reports.')
# Return success value
if not clean_run:
return SLAPGRID_FAIL
return SLAPGRID_SUCCESS
def destroy(self):
"""Destroys the partition and makes it available for subsequent use."
"""
self.logger.info("Destroying Computer Partition %s..."
% self.computer_partition.getId())
self.createRetentionLockDate()
if not self.checkRetentionIsAuthorized():
return False
# Launches "destroy" binary if exists
destroy_executable_location = os.path.join(self.instance_path, 'sbin',
'destroy')
if os.path.exists(destroy_executable_location):
uid, gid = self.getUserGroupId()
self.logger.debug('Invoking %r' % destroy_executable_location)
process_handler = SlapPopen([destroy_executable_location],
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=self.instance_path,
env=getCleanEnvironment(logger=self.logger,
home_path=pwd.getpwuid(uid).pw_dir),
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
logger=self.logger)
if process_handler.returncode is None or process_handler.returncode != 0:
message = 'Failed to destroy Computer Partition in %r.' % \
self.instance_path
self.logger.error(message)
raise subprocess.CalledProcessError(message, process_handler.output)
# Manually cleans what remains
try:
for f in [self.key_file, self.cert_file]:
if f:
if os.path.exists(f):
os.unlink(f)
# better to manually remove symlinks because rmtree might choke on them
sr_symlink = os.path.join(self.instance_path, 'software_release')
if os.path.islink(sr_symlink):
os.unlink(sr_symlink)
data_base_link = os.path.join(self.instance_path, CP_STORAGE_FOLDER_NAME)
if self.instance_storage_home and os.path.exists(data_base_link) and \
os.path.isdir(data_base_link):
for filename in os.listdir(data_base_link):
data_symlink = os.path.join(data_base_link, filename)
partition_data_path = os.path.join(self.instance_storage_home,
filename, self.partition_id)
if os.path.lexists(data_symlink):
os.unlink(data_symlink)
if os.path.exists(partition_data_path):
self.cleanupFolder(partition_data_path)
self.cleanupFolder(self.instance_path)
# Cleanup all Data storage location of this partition
if os.path.exists(self.supervisord_partition_configuration_path):
os.remove(self.supervisord_partition_configuration_path)
self.updateSupervisor()
except IOError as exc:
raise IOError("I/O error while freeing partition (%s): %s" % (self.instance_path, exc))
return True
def install(self):
""" Creates configuration file from template in software_path, then
installs the software partition with the help of buildout
"""
self.logger.info("Installing Computer Partition %s..."
% self.computer_partition.getId())
self.check_free_space()
# Checks existence and permissions of Partition directory
# Note : Partitions have to be created and configured before running slapgrid
if not os.path.isdir(self.instance_path):
raise PathDoesNotExistError('Please create partition directory %s'
% self.instance_path)
sr_symlink = os.path.join(self.instance_path, 'software_release')
self.updateSymlink(sr_symlink, self.software_path)
instance_stat_info = os.stat(self.instance_path)
permission = stat.S_IMODE(instance_stat_info.st_mode)
if permission != REQUIRED_COMPUTER_PARTITION_PERMISSION:
raise WrongPermissionError('Wrong permissions in %s: actual '
'permissions are: 0%o, wanted are 0%o' %
(self.instance_path, permission,
REQUIRED_COMPUTER_PARTITION_PERMISSION))
os.environ = getCleanEnvironment(logger=self.logger,
home_path=pwd.getpwuid(instance_stat_info.st_uid).pw_dir)
# Check that Software Release directory is present
if not os.path.exists(self.software_path):
# XXX What should it raise?
raise IOError('Software Release %s is not present on system.\n'
'Cannot deploy instance.' % self.software_release_url)
# Generate buildout instance profile from template in Software Release
template_location = os.path.join(self.software_path, 'instance.cfg')
if not os.path.exists(template_location):
# Backward compatibility: "instance.cfg" file was named "template.cfg".
if os.path.exists(os.path.join(self.software_path, 'template.cfg')):
template_location = os.path.join(self.software_path, 'template.cfg')
else:
# No template: Software Release is either inconsistent or not correctly installed.
# XXX What should it raise?
raise IOError('Software Release %s is not correctly installed.\nMissing file: %s' % (
self.software_release_url, template_location))
config_location = os.path.join(self.instance_path, 'buildout.cfg')
self.logger.debug("Copying %r to %r" % (template_location, config_location))
shutil.copy(template_location, config_location)
# fill generated buildout with additional information
buildout_text = open(config_location).read()
buildout_text += '\n\n' + pkg_resources.resource_string(__name__,
'templates/buildout-tail.cfg.in') % {
'computer_id': self.computer_id,
'partition_id': self.partition_id,
'server_url': self.server_url,
'software_release_url': self.software_release_url,
'key_file': self.key_file,
'cert_file': self.cert_file,
'storage_home': self.instance_storage_home,
'global_ipv4_network_prefix': self.ipv4_global_network,
}
open(config_location, 'w').write(buildout_text)
os.chmod(config_location, 0o640)
# Try to find the best possible buildout:
# *) if software_root/bin/bootstrap exists use this one to bootstrap
# locally
# *) as last resort fallback to buildout binary from software_path
bootstrap_candidate_dir = os.path.abspath(os.path.join(self.software_path,
'bin'))
if os.path.isdir(bootstrap_candidate_dir):
bootstrap_candidate_list = [q for q in os.listdir(bootstrap_candidate_dir)
if q.startswith('bootstrap')]
else:
bootstrap_candidate_list = []
uid, gid = self.getUserGroupId()
os.chown(config_location, -1, int(gid))
if len(bootstrap_candidate_list) == 0:
buildout_binary = os.path.join(self.software_path, 'bin', 'buildout')
self.logger.info("Falling back to default buildout %r" %
buildout_binary)
else:
if len(bootstrap_candidate_list) != 1:
raise ValueError('More than one bootstrap candidate found.')
# Reads uid/gid of path, launches buildout with thoses privileges
bootstrap_file = os.path.abspath(os.path.join(bootstrap_candidate_dir,
bootstrap_candidate_list[0]))
first_line = open(bootstrap_file, 'r').readline()
invocation_list = []
if first_line.startswith('#!'):
invocation_list = first_line[2:].split()
invocation_list.append(bootstrap_file)
self.logger.debug('Invoking %r in %r' % (' '.join(invocation_list),
self.instance_path))
process_handler = SlapPopen(invocation_list,
preexec_fn=lambda: dropPrivileges(uid, gid, logger=self.logger),
cwd=self.instance_path,
env=getCleanEnvironment(logger=self.logger,
home_path=pwd.getpwuid(uid).pw_dir),
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
logger=self.logger)
if process_handler.returncode is None or process_handler.returncode != 0:
message = 'Failed to bootstrap buildout in %r.' % (self.instance_path)
self.logger.error(message)
raise BuildoutFailedError('%s:\n%s\n' % (message, process_handler.output))
buildout_binary = os.path.join(self.instance_path, 'sbin', 'buildout')
if not os.path.exists(buildout_binary):
# use own buildout generation
utils.bootstrapBuildout(path=self.instance_path,
buildout=self.buildout,
logger=self.logger,
additional_buildout_parameter_list=
['buildout:bin-directory=%s' %
os.path.join(self.instance_path, 'sbin')])
buildout_binary = os.path.join(self.instance_path, 'sbin', 'buildout')
# Launches buildout
utils.launchBuildout(path=self.instance_path,
buildout_binary=buildout_binary,
logger=self.logger)
self.generateSupervisorConfigurationFile()
self.createRetentionLockDelay()