class DeploymentRPCInterface:
def __init__(self, supervisord, **config):
self.supervisord = supervisord
self.global_config = config
self.supervisor_rpcinterface = SupervisorNamespaceRPCInterface(supervisord)
self.package_server = config.get('package_server')
self.download_package_uri = config.get('download_package_uri')
self.get_latest_package_info_uri = config.get('get_latest_package_info_uri')
def get_run_dir(self, service, cluster, job, instance_id=-1):
'''
Get the run directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's run root directory
'''
app_root = self.global_config.get('app_root', DEFAULT_APP_ROOT)
if instance_id == -1:
return '%s/%s/%s/%s' % (app_root, service, cluster, job)
else:
return '%s/%s/%s/%s/%s' % (app_root, service, cluster, job, instance_id)
def get_log_dir(self, service, cluster, job, instance_id=-1):
'''
Get the log directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's log root directory
'''
log_root = self.global_config.get('log_root', DEFAULT_LOG_ROOT)
if instance_id == -1:
return '%s/%s/%s/%s' % (log_root, service, cluster, job)
else:
return '%s/%s/%s/%s/%s' % (log_root, service, cluster, job, instance_id)
def get_stdout_dir(self, service, cluster, job, instance_id=-1):
'''
Get the stdout directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's log root directory
'''
run_dir = self.get_run_dir(service, cluster, job, instance_id)
return '%s/stdout' % run_dir
def get_available_data_dirs(self, service, cluster, job, instance_id=-1):
'''
Get all the available data directories that the specified job may use
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return list all the available data root directories
'''
data_dirs = self.global_config.get('data_dirs', DEFAULT_DATA_DIRS)
if instance_id == -1:
return ['%s/%s/%s/%s' % (data_dir, service, cluster, job)
for data_dir in data_dirs.split(',')
]
else:
return ['%s/%s/%s/%s/%s' % (data_dir, service, cluster, job, instance_id)
for data_dir in data_dirs.split(',')
]
def get_data_dirs(self, service, cluster, job, instance_id=-1):
'''
Get all the data directories of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return list the job's data root directories
'''
file_name = '%s/%s' % (self.get_run_dir(service, cluster, job, instance_id),
JOB_RUN_CONFIG)
if not os.path.exists(file_name):
return 'You should bootstrapped the job first'
data_dirs = self.get_available_data_dirs(service, cluster, job, instance_id)
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_name])
data_dir_indexes = run_config.get('run_info', 'data_dir_indexes')
job_data_dirs = []
for i in data_dir_indexes.split(','):
job_data_dirs.append(data_dirs[int(i)])
return job_data_dirs
def get_package_dir(self, service, cluster, job, instance_id=-1):
'''
Get the current package directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's package root directory(symbol link)
'''
return '%s/package' % self.get_run_dir(service, cluster, job, instance_id)
def get_real_package_dir(self, service, cluster, job, instance_id=-1):
'''
Get the current package directory real path of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's package root directory(real path)
'''
return os.readlink(self.get_package_dir(service, cluster, job, instance_id))
def get_current_package_dir(self, service, cluster):
'''
Get the currently used package directory of the specified service
@param service the service name
@param cluster the cluster name
@return string the currently used package directory
'''
package_root = self.global_config.get('package_root')
return '%s/%s/%s/current' % (package_root, service, cluster)
def get_cleanup_token(self, service, cluster, job, instance_id=-1):
'''
Get the token used to do cleanuping
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@return string the job's cleanup token
'''
file_name = '%s/%s' % (self.get_run_dir(service, cluster, job, instance_id),
JOB_RUN_CONFIG)
if not os.path.exists(file_name):
return 'You should bootstrap the job first'
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_name])
return run_config.get('run_info', 'cleanup_token')
def bootstrap(self, service, cluster, job, config_dict, instance_id=-1):
'''
Bootstrap the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict must contain the following item:
1. artifact
2. bootstrap.sh
3. if any config files are needed, just put it in 'config_files' item
config_dict can also contain the following optional items:
1. cleanup_token: if this token is specified, user should supply
the token to do cleanup
2. package_name: package_name, revision, timestamp should be specified
simultaneously, otherwise will be ignored
3. revision
4. timestamp
5. data_dir_indexes: if this is not specified, the first data_dir is
used by default
6. force_update
This is an example:
config_dict = {
'artifact': 'hadoop',
'bootstrap.sh': $bootstrap_file_content,
'config_files': {
'core-site.xml': $core_site_xml_content,
...
},
}
'''
return self._do_bootstrap(service, cluster, job, instance_id, **config_dict)
def start(self, service, cluster, job, config_dict, instance_id=-1):
'''
Start the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict must contain the following item:
1. start.sh
2. artifact
3. if any config files are needed, just put it in 'config_files' item
config_dict can also contain the following optional items:
1. http_url: the server's http service url
2. package_name: package_name, revision, timestamp should be specified
simultaneously, otherwise will be ignored
3. revision
4. timestamp
5. force_update
This is an example:
config_dict = {
'start.sh': $start_file_content,
'artifact': hadoop,
'config_files': {
'core-site.xml': $core_site_xml_content,
...
},
'http_url': 'http://10.235.3.67:11201',
}
'''
return self._do_start(service, cluster, job, instance_id, **config_dict)
def stop(self, service, cluster, job, config_dict, instance_id=-1):
'''
Stop the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict is not used currently, reserved for extendibility
'''
return self._do_stop(service, cluster, job, instance_id, **config_dict)
def cleanup(self, service, cluster, job, config_dict, instance_id=-1):
'''
Cleanup the specified job's data/log directories
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict may contain the following item:
1. cleanup_token: [optional] token used to do verification
2. cleanup.sh: [optional] script used to do cleanuping
This is an example:
config_dict = {
'cleanup_token': '550e8400-e29b-41d4-a716-446655440000',
'cleanup.sh': $cleanup_script,
}
'''
return self._do_cleanup(service, cluster, job, instance_id, **config_dict)
def show(self, service, cluster, job, config_dict, instance_id=-1):
'''
Get the specified job's current status
@param service the server name
@param cluster the cluster name
@param job the job name
@param instance_id the instance id
@param config_dict the config information dictionary
@return string the process status
Possible values of process status:
RUNNING STARTING BACKOFF STOPPING EXITED FATAL UNKNOWN
Note: config_dict is not used currently, reserved for extendibility
'''
return self._do_show(service, cluster, job, instance_id, **config_dict)
def read_file(self, file_path):
'''
Read the file with the given file path on server
@param file_path the name of file to read
'''
with open(file_path) as fi:
return fi.read()
def write_text_to_file(self, file_path, content):
'''
Write content to the file with the given file path on server
@param file_path the name of file to write
@param content the content to write
'''
with open(file_path, 'w') as fi:
fi.write(content)
return 'OK'
def append_text_to_file(self, file_path, content):
'''
Append content to the file with the given file path on server
@param file_path the name of file to append
@param content the content to append
'''
with open(file_path, 'a') as fi:
fi.write(content)
return 'OK'
def _get_package_uri(self, artifact, revision, timestamp, package_name):
return '%s/%s/%s/%s-%s/%s' % (self.package_server,
self.download_package_uri, artifact,
revision, timestamp, package_name)
def _get_query_latest_package_info_uri(self, artifact):
return '%s/%s/?artifact=%s' % (self.package_server,
self.get_latest_package_info_uri, artifact)
def _downlowd_package(self, uri, dest_file):
data_file = urllib2.urlopen(uri, None, 30)
if not os.path.exists(os.path.dirname(dest_file)):
os.makedirs(os.path.dirname(dest_file))
fp = open(dest_file, 'wb')
fp.write(data_file.read())
fp.close()
data_file.close()
def _write_file(self, file_path, file_content):
fp = open(file_path, 'wb')
fp.write(file_content)
fp.close()
def _write_config_files(self, run_dir, **config_dict):
for file_name, content in config_dict.iteritems():
file_path = '%s/%s' % (run_dir, file_name)
if os.path.exists(file_path):
os.remove(file_path)
self._write_file(file_path, content)
def _get_process_name(self, service, cluster, job, instance_id):
if instance_id == -1:
return '%s--%s--%s' % (service, cluster, job)
else:
return '%s--%s--%s%d' % (service, cluster, job, instance_id)
def _cleanup_dir(self, path):
cmd = 'rm -rf %s/*' % path
subprocess.check_call(cmd, shell=True)
def _check_dir_empty(self, path):
if not os.path.exists(path):
return True
lists = os.listdir(path)
return len(lists) == 0
def _check_bootstrapped(self, service, cluster, job, instance_id):
run_dir = self.get_run_dir(service, cluster, job, instance_id)
return os.path.exists('%s/%s' % (run_dir, JOB_RUN_CONFIG))
def _get_latest_package_info(self, artifact):
uri = self._get_query_latest_package_info_uri(artifact)
info_fp = urllib2.urlopen(uri, None, 30)
info = info_fp.read()
if info and info.startswith('{'):
info_dict = eval(info)
info_fp.close()
return info_dict
else:
info_fp.close()
return None
def _make_package_dir(self, artifact, service, cluster, job, instance_id,
revision, timestamp, package_name):
# Check if the tarball is already downloaded, if not, download it
package_path = '%s/%s/%s/%s-%s/%s' % (self.global_config.get('package_root'),
service, cluster, revision, timestamp, package_name)
if not os.path.exists(package_path):
self._downlowd_package(
self._get_package_uri(artifact, revision, timestamp, package_name),
package_path)
# Unpack the tarball
package_dir = package_path[0: len(package_path) - len('.tar.gz')]
if os.path.exists(package_dir):
cmd = ['rm', '-rf', package_dir]
subprocess.check_call(cmd)
cmd = ['tar', '-zxf', package_path, '-C', os.path.dirname(package_dir)]
subprocess.check_call(cmd)
# Link the package dir to the 'current'
current_dir = self.get_current_package_dir(service, cluster)
if os.path.lexists(current_dir):
os.unlink(current_dir)
os.symlink(package_dir, current_dir)
# Link the package dir to the run dir
symbol_package_dir = self.get_package_dir(service, cluster, job, instance_id)
if os.path.lexists(symbol_package_dir):
os.unlink(symbol_package_dir)
os.symlink(package_dir, symbol_package_dir)
return package_dir
def _update_run_cfg(self, file_path, section, key, value):
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_path])
run_config.set(section, key, value)
fp = open(file_path, 'w')
run_config.write(fp)
fp.close()
def _prepare_run_env(self, service, cluster, job, instance_id, **config_dict):
artifact = config_dict.get('artifact')
if not artifact:
return 'Invalid config_dict: can\'t find artifact'
# Create run dirs
run_dir = self.get_run_dir(service, cluster, job, instance_id)
if not os.path.exists(run_dir):
os.makedirs(run_dir)
# Create stdout dir
stdout_dir = self.get_stdout_dir(service, cluster, job, instance_id)
if not os.path.exists(stdout_dir):
os.makedirs(stdout_dir)
# Create and link log dir to the run dir
log_dir = self.get_log_dir(service, cluster, job, instance_id)
if os.path.exists(log_dir):
if not self._check_dir_empty(log_dir):
return 'The log dir %s is not empty, please do cleanup first' % log_dir
else:
os.makedirs(log_dir)
symbol_log_dir = '%s/log' % run_dir
if not os.path.exists(symbol_log_dir):
os.symlink(log_dir, symbol_log_dir)
# Create and link data dirs to the run dir
data_dirs = self.global_config.get('data_dirs', DEFAULT_DATA_DIRS).split(',')
data_dir_indexes = (config_dict.get('data_dir_indexes') or '0')
for i in data_dir_indexes.split(','):
if instance_id == -1:
data_dir = '%s/%s/%s/%s' % (data_dirs[int(i)], service, cluster, job)
else:
data_dir = '%s/%s/%s/%s/%s' % (data_dirs[int(i)], service, cluster, job, instance_id)
if os.path.exists(data_dir):
if not self._check_dir_empty(data_dir):
return 'The data dir %s is not empty, please do cleanup first' % data_dir
else:
os.makedirs(data_dir)
symbol_data_dir = '%s/%s' % (run_dir, os.path.basename(data_dirs[int(i)]))
if not os.path.exists(symbol_data_dir):
os.symlink(data_dir, symbol_data_dir)
# Check the package information
force_update = config_dict.get('force_update', False)
if force_update:
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
else:
package_name = config_dict.get('package_name')
revision = config_dict.get('revision')
timestamp = config_dict.get('timestamp')
if not (package_name and revision and timestamp):
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
if not (package_name and revision and timestamp):
return 'No package found on package server of %s' % artifact
# Write the job's run.cfg
package_dir = self._make_package_dir(artifact, service, cluster, job,
instance_id, revision, timestamp, package_name)
cleanup_token = config_dict.get('cleanup_token', str())
run_config = ConfigParser.SafeConfigParser()
run_config.add_section('run_info')
run_config.set('run_info', 'cleanup_token', cleanup_token)
run_config.set('run_info', 'data_dir_indexes', data_dir_indexes)
run_config.set('run_info', 'run_dir', run_dir)
run_config.set('run_info', 'log_dir', log_dir)
run_config.set('run_info', 'package_dir', package_dir)
fp = open('%s/%s' % (run_dir, JOB_RUN_CONFIG), 'w')
run_config.write(fp)
fp.close()
return SUCCESS_STATUS
def _do_bootstrap(self, service, cluster, job, instance_id, **config_dict):
# prepare run dir
message = self._prepare_run_env(service, cluster, job, instance_id, **config_dict)
if message != SUCCESS_STATUS:
return message
# Write other config files to local disk
config_files = config_dict.get('config_files')
service_root = self.get_run_dir(service, cluster, job, instance_id)
if config_files:
self._write_config_files(service_root, **config_files)
# Do bootstraping
bootstrap_sh = config_dict.get('bootstrap.sh')
if bootstrap_sh:
self._write_file('%s/bootstrap.sh' % service_root, bootstrap_sh)
cmd = ['/bin/bash', '%s/bootstrap.sh' % service_root]
subprocess.call(cmd)
return SUCCESS_STATUS
def _do_start(self, service, cluster, job, instance_id, **config_dict):
artifact = config_dict.get('artifact')
if not artifact:
return 'Inval config_dict: can\'t find artifact'
if not self._check_bootstrapped(service, cluster, job, instance_id):
return "You should bootstrap the job first"
# Check if need update the package
force_update = config_dict.get('force_update', False)
if force_update:
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
else:
package_name = config_dict.get('package_name')
revision = config_dict.get('revision')
timestamp = config_dict.get('timestamp')
if (package_name and revision and timestamp):
package_path = '%s/%s/%s-%s/%s' % (
self.global_config.get('package_root'),
artifact, revision, timestamp, package_name)
if not os.path.exists(package_path):
self._downlowd_package(
self._get_package_uri(artifact, revision, timestamp, package_name),
package_path)
package_dir = self._make_package_dir(artifact, service, cluster, job,
instance_id, revision, timestamp, package_name)
run_cfg = '%s/%s' % (self.get_run_dir(service, cluster, job, instance_id),
JOB_RUN_CONFIG)
self._update_run_cfg(run_cfg, 'run_info', 'package_dir', package_dir)
# Write the start script to local disk
start_sh = config_dict.get('start.sh')
service_root = self.get_run_dir(service, cluster, job, instance_id)
if not start_sh and not os.path.exists('%s/start.sh' % service_root):
return 'No start script found'
elif start_sh:
self._write_file('%s/start.sh' % service_root, start_sh)
# Write other config files to local disk
config_files = config_dict.get('config_files')
if config_files:
self._write_config_files(service_root, **config_files)
# Write supervisor config
http_url = config_dict.get('http_url', '')
process_name = self._get_process_name(service, cluster, job, instance_id)
job_config = ConfigParser.SafeConfigParser()
section = 'program:%s' % process_name
job_config.add_section(section)
job_config.set(section, 'command', '/bin/bash %s/start.sh' % service_root)
job_config.set(section, 'process_name', process_name)
job_config.set(section, 'directory', service_root)
job_config.set(section, 'http_url', http_url)
# Process will be unconditionally restarted when it exits, without regard
# to its exit code
job_config.set(section, 'autorestart', 'true')
job_config.set(section, 'exitcodes', str(DEFAULT_EXPECTED_EXIT_CODE))
# Process will NOT be automatically started when supervisor restart.
job_config.set(section, 'autostart', 'false')
fp = open('%s/%s/%s.cfg' % (os.getcwd(), CONFIG_PATH, process_name), 'wb')
job_config.write(fp)
fp.close()
# Start the job
self.supervisor_rpcinterface.reloadConfig()
try:
self.supervisor_rpcinterface.addProcessGroup(process_name)
except RPCError, e:
if e.code != Faults.ALREADY_ADDED:
raise e
self.supervisor_rpcinterface.startProcess(process_name)()
return SUCCESS_STATUS
class DeploymentRPCInterface:
def __init__(self, supervisord, **config):
self.supervisord = supervisord
self.global_config = config
self.supervisor_rpcinterface = SupervisorNamespaceRPCInterface(
supervisord)
self.package_server = config.get('package_server')
self.download_package_uri = config.get('download_package_uri')
self.get_latest_package_info_uri = config.get(
'get_latest_package_info_uri')
def get_run_dir(self, service, cluster, job):
'''
Get the run directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's run root directory
'''
app_root = self.global_config.get('app_root', DEFAULT_APP_ROOT)
return '%s/%s/%s/%s' % (app_root, service, cluster, job)
def get_log_dir(self, service, cluster, job):
'''
Get the log directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's log root directory
'''
log_root = self.global_config.get('log_root', DEFAULT_LOG_ROOT)
return '%s/%s/%s/%s' % (log_root, service, cluster, job)
def get_stdout_dir(self, service, cluster, job):
'''
Get the stdout directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's log root directory
'''
run_dir = self.get_run_dir(service, cluster, job)
return '%s/stdout' % run_dir
def get_available_data_dirs(self, service, cluster, job):
'''
Get all the available data directories that the specified job may use
@param service the server name
@param cluster the cluster name
@param job the job name
@return list all the available data root directories
'''
data_dirs = self.global_config.get('data_dirs', DEFAULT_DATA_DIRS)
return [
'%s/%s/%s/%s' % (data_dir, service, cluster, job)
for data_dir in data_dirs.split(',')
]
def get_data_dirs(self, service, cluster, job):
'''
Get all the data directories of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return list the job's data root directories
'''
file_name = '%s/%s' % (self.get_run_dir(service, cluster,
job), JOB_RUN_CONFIG)
if not os.path.exists(file_name):
return 'You should bootstrapped the job first'
data_dirs = self.get_available_data_dirs(service, cluster, job)
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_name])
data_dir_indexes = run_config.get('run_info', 'data_dir_indexes')
job_data_dirs = []
for i in data_dir_indexes.split(','):
job_data_dirs.append(data_dirs[int(i)])
return job_data_dirs
def get_package_dir(self, service, cluster, job):
'''
Get the current package directory of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's package root directory(symbol link)
'''
return '%s/package' % self.get_run_dir(service, cluster, job)
def get_real_package_dir(self, service, cluster, job):
'''
Get the current package directory real path of the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's package root directory(real path)
'''
return os.readlink(self.get_package_dir(service, cluster, job))
def get_current_package_dir(self, service, cluster):
'''
Get the currently used package directory of the specified service
@param service the service name
@param cluster the cluster name
@return string the currently used package directory
'''
package_root = self.global_config.get('package_root')
return '%s/%s/%s/current' % (package_root, service, cluster)
def get_cleanup_token(self, service, cluster, job):
'''
Get the token used to do cleanuping
@param service the server name
@param cluster the cluster name
@param job the job name
@return string the job's cleanup token
'''
file_name = '%s/%s' % (self.get_run_dir(service, cluster,
job), JOB_RUN_CONFIG)
if not os.path.exists(file_name):
return 'You should bootstrap the job first'
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_name])
return run_config.get('run_info', 'cleanup_token')
def bootstrap(self, service, cluster, job, config_dict):
'''
Bootstrap the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict must contain the following item:
1. artifact
2. bootstrap.sh
3. if any config files are needed, just put it in 'config_files' item
config_dict can also contain the following optional items:
1. cleanup_token: if this token is specified, user should supply
the token to do cleanup
2. package_name: package_name, revision, timestamp should be specified
simultaneously, otherwise will be ignored
3. revision
4. timestamp
5. data_dir_indexes: if this is not specified, the first data_dir is
used by default
6. force_update
This is an example:
config_dict = {
'artifact': 'hadoop',
'bootstrap.sh': $bootstrap_file_content,
'config_files': {
'core-site.xml': $core_site_xml_content,
...
},
}
'''
return self._do_bootstrap(service, cluster, job, **config_dict)
def start(self, service, cluster, job, config_dict):
'''
Start the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict must contain the following item:
1. start.sh
2. artifact
3. if any config files are needed, just put it in 'config_files' item
config_dict can also contain the following optional items:
1. http_url: the server's http service url
2. package_name: package_name, revision, timestamp should be specified
simultaneously, otherwise will be ignored
3. revision
4. timestamp
5. force_update
This is an example:
config_dict = {
'start.sh': $start_file_content,
'artifact': hadoop,
'config_files': {
'core-site.xml': $core_site_xml_content,
...
},
'http_url': 'http://10.235.3.67:11201',
}
'''
return self._do_start(service, cluster, job, **config_dict)
def stop(self, service, cluster, job, config_dict):
'''
Stop the specified job
@param service the server name
@param cluster the cluster name
@param job the job name
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict is not used currently, reserved for extendibility
'''
return self._do_stop(service, cluster, job, **config_dict)
def cleanup(self, service, cluster, job, config_dict):
'''
Cleanup the specified job's data/log directories
@param service the server name
@param cluster the cluster name
@param job the job name
@param config_dict the config information dictionary
@return string 'OK' on success, otherwise, the error message
Note: config_dict may contain the following item:
1. cleanup_token: [optional] token used to do verification
2. cleanup.sh: [optional] script used to do cleanuping
This is an example:
config_dict = {
'cleanup_token': '550e8400-e29b-41d4-a716-446655440000',
'cleanup.sh': $cleanup_script,
}
'''
return self._do_cleanup(service, cluster, job, **config_dict)
def show(self, service, cluster, job, config_dict):
'''
Get the specified job's current status
@param service the server name
@param cluster the cluster name
@param job the job name
@param config_dict the config information dictionary
@return string the process status
Possible values of process status:
RUNNING STARTING BACKOFF STOPPING EXITED FATAL UNKNOWN
Note: config_dict is not used currently, reserved for extendibility
'''
return self._do_show(service, cluster, job, **config_dict)
def _get_package_uri(self, artifact, revision, timestamp, package_name):
return '%s/%s/%s/%s-%s/%s' % (self.package_server,
self.download_package_uri, artifact,
revision, timestamp, package_name)
def _get_query_latest_package_info_uri(self, artifact):
return '%s/%s/?artifact=%s' % (
self.package_server, self.get_latest_package_info_uri, artifact)
def _downlowd_package(self, uri, dest_file):
data_file = urllib2.urlopen(uri, None, 30)
if not os.path.exists(os.path.dirname(dest_file)):
os.makedirs(os.path.dirname(dest_file))
fp = open(dest_file, 'wb')
fp.write(data_file.read())
fp.close()
data_file.close()
def _write_file(self, file_path, file_content):
fp = open(file_path, 'wb')
fp.write(file_content)
fp.close()
def _write_config_files(self, run_dir, **config_dict):
for file_name, content in config_dict.iteritems():
file_path = '%s/%s' % (run_dir, file_name)
if os.path.exists(file_path):
os.remove(file_path)
self._write_file(file_path, content)
def _get_process_name(self, service, cluster, job):
return '%s--%s--%s' % (service, cluster, job)
def _cleanup_dir(self, path):
cmd = 'rm -rf %s/*' % path
subprocess.check_call(cmd, shell=True)
def _check_dir_empty(self, path):
if not os.path.exists(path):
return True
lists = os.listdir(path)
return len(lists) == 0
def _check_bootstrapped(self, service, cluster, job):
run_dir = self.get_run_dir(service, cluster, job)
return os.path.exists('%s/%s' % (run_dir, JOB_RUN_CONFIG))
def _get_latest_package_info(self, artifact):
uri = self._get_query_latest_package_info_uri(artifact)
info_fp = urllib2.urlopen(uri, None, 30)
info = info_fp.read()
if info and info.startswith('{'):
info_dict = eval(info)
info_fp.close()
return info_dict
else:
info_fp.close()
return None
def _make_package_dir(self, artifact, service, cluster, job, revision,
timestamp, package_name):
# Check if the tarball is already downloaded, if not, download it
package_path = '%s/%s/%s/%s-%s/%s' % (
self.global_config.get('package_root'), service, cluster, revision,
timestamp, package_name)
if not os.path.exists(package_path):
self._downlowd_package(
self._get_package_uri(artifact, revision, timestamp,
package_name), package_path)
# Unpack the tarball
package_dir = package_path[0:len(package_path) - len('.tar.gz')]
if os.path.exists(package_dir):
cmd = ['rm', '-rf', package_dir]
subprocess.check_call(cmd)
cmd = ['tar', '-zxf', package_path, '-C', os.path.dirname(package_dir)]
subprocess.check_call(cmd)
# Link the package dir to the 'current'
current_dir = self.get_current_package_dir(service, cluster)
if os.path.lexists(current_dir):
os.unlink(current_dir)
os.symlink(package_dir, current_dir)
# Link the package dir to the run dir
symbol_package_dir = self.get_package_dir(service, cluster, job)
if os.path.lexists(symbol_package_dir):
os.unlink(symbol_package_dir)
os.symlink(package_dir, symbol_package_dir)
return package_dir
def _update_run_cfg(self, file_path, section, key, value):
run_config = ConfigParser.SafeConfigParser()
run_config.read([file_path])
run_config.set(section, key, value)
fp = open(file_path, 'w')
run_config.write(fp)
fp.close()
def _prepare_run_env(self, service, cluster, job, **config_dict):
artifact = config_dict.get('artifact')
if not artifact:
return 'Invalid config_dict: can\'t find artifact'
# Create run dirs
run_dir = self.get_run_dir(service, cluster, job)
if not os.path.exists(run_dir):
os.makedirs(run_dir)
# Create stdout dir
stdout_dir = self.get_stdout_dir(service, cluster, job)
if not os.path.exists(stdout_dir):
os.makedirs(stdout_dir)
# Create and link log dir to the run dir
log_dir = self.get_log_dir(service, cluster, job)
if os.path.exists(log_dir):
if not self._check_dir_empty(log_dir):
return 'The log dir %s is not empty, please do cleanup first' % log_dir
else:
os.makedirs(log_dir)
symbol_log_dir = '%s/log' % run_dir
if not os.path.exists(symbol_log_dir):
os.symlink(log_dir, symbol_log_dir)
# Create and link data dirs to the run dir
data_dirs = self.global_config.get('data_dirs',
DEFAULT_DATA_DIRS).split(',')
data_dir_indexes = (config_dict.get('data_dir_indexes') or '0')
for i in data_dir_indexes.split(','):
data_dir = '%s/%s/%s/%s' % (data_dirs[int(i)], service, cluster,
job)
if os.path.exists(data_dir):
if not self._check_dir_empty(data_dir):
return 'The data dir %s is not empty, please do cleanup first' % data_dir
else:
os.makedirs(data_dir)
symbol_data_dir = '%s/%s' % (run_dir,
os.path.basename(data_dirs[int(i)]))
if not os.path.exists(symbol_data_dir):
os.symlink(data_dir, symbol_data_dir)
# Check the package information
force_update = config_dict.get('force_update', False)
if force_update:
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
else:
package_name = config_dict.get('package_name')
revision = config_dict.get('revision')
timestamp = config_dict.get('timestamp')
if not (package_name and revision and timestamp):
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
if not (package_name and revision and timestamp):
return 'No package found on package server of %s' % artifact
# Write the job's run.cfg
package_dir = self._make_package_dir(artifact, service, cluster, job,
revision, timestamp, package_name)
cleanup_token = config_dict.get('cleanup_token', str())
run_config = ConfigParser.SafeConfigParser()
run_config.add_section('run_info')
run_config.set('run_info', 'cleanup_token', cleanup_token)
run_config.set('run_info', 'data_dir_indexes', data_dir_indexes)
run_config.set('run_info', 'run_dir', run_dir)
run_config.set('run_info', 'log_dir', log_dir)
run_config.set('run_info', 'package_dir', package_dir)
fp = open('%s/%s' % (run_dir, JOB_RUN_CONFIG), 'w')
run_config.write(fp)
fp.close()
return SUCCESS_STATUS
def _do_bootstrap(self, service, cluster, job, **config_dict):
# prepare run dir
message = self._prepare_run_env(service, cluster, job, **config_dict)
if message != SUCCESS_STATUS:
return message
# Write other config files to local disk
config_files = config_dict.get('config_files')
service_root = self.get_run_dir(service, cluster, job)
if config_files:
self._write_config_files(service_root, **config_files)
# Do bootstraping
bootstrap_sh = config_dict.get('bootstrap.sh')
if bootstrap_sh:
self._write_file('%s/bootstrap.sh' % service_root, bootstrap_sh)
cmd = ['/bin/bash', '%s/bootstrap.sh' % service_root]
subprocess.call(cmd)
return SUCCESS_STATUS
def _do_start(self, service, cluster, job, **config_dict):
artifact = config_dict.get('artifact')
if not artifact:
return 'Inval config_dict: can\'t find artifact'
if not self._check_bootstrapped(service, cluster, job):
return "You should bootstrap the job first"
# Check if need update the package
force_update = config_dict.get('force_update', False)
if force_update:
package_info = self._get_latest_package_info(artifact)
if package_info:
package_name = package_info.get('package_name')
revision = package_info.get('revision')
timestamp = package_info.get('timestamp')
else:
package_name = config_dict.get('package_name')
revision = config_dict.get('revision')
timestamp = config_dict.get('timestamp')
if (package_name and revision and timestamp):
package_path = '%s/%s/%s-%s/%s' % (self.global_config.get(
'package_root'), artifact, revision, timestamp, package_name)
if not os.path.exists(package_path):
self._downlowd_package(
self._get_package_uri(artifact, revision, timestamp,
package_name), package_path)
package_dir = self._make_package_dir(artifact, service, cluster,
job, revision, timestamp,
package_name)
run_cfg = '%s/%s' % (self.get_run_dir(service, cluster,
job), JOB_RUN_CONFIG)
self._update_run_cfg(run_cfg, 'run_info', 'package_dir',
package_dir)
# Write the start script to local disk
start_sh = config_dict.get('start.sh')
service_root = self.get_run_dir(service, cluster, job)
if not start_sh and not os.path.exists('%s/start.sh' % service_root):
return 'No start script found'
elif start_sh:
self._write_file('%s/start.sh' % service_root, start_sh)
# Write other config files to local disk
config_files = config_dict.get('config_files')
if config_files:
self._write_config_files(service_root, **config_files)
# Write supervisor config
http_url = config_dict.get('http_url', '')
process_name = self._get_process_name(service, cluster, job)
job_config = ConfigParser.SafeConfigParser()
section = 'program:%s' % process_name
job_config.add_section(section)
job_config.set(section, 'command',
'/bin/bash %s/start.sh' % service_root)
job_config.set(section, 'process_name', process_name)
job_config.set(section, 'directory', service_root)
job_config.set(section, 'http_url', http_url)
# Process will be unconditionally restarted when it exits, without regard
# to its exit code
job_config.set(section, 'autorestart', 'true')
job_config.set(section, 'exitcodes', str(DEFAULT_EXPECTED_EXIT_CODE))
# Process will NOT be automatically started when supervisor restart.
job_config.set(section, 'autostart', 'false')
fp = open('%s/%s/%s.cfg' % (os.getcwd(), CONFIG_PATH, process_name),
'wb')
job_config.write(fp)
fp.close()
# Start the job
self.supervisor_rpcinterface.reloadConfig()
try:
self.supervisor_rpcinterface.addProcessGroup(process_name)
except RPCError, e:
if e.code != Faults.ALREADY_ADDED:
raise e
self.supervisor_rpcinterface.startProcess(process_name)()
return SUCCESS_STATUS
