def test_lint_profile_configs():
expected_profile_0 = Profile(name='profile_0',
metrics=['node_logged_in_users', 'node_open_file_descriptors'],
logger=test_logger)
expected_profile_1 = Profile(name='profile_1',
metrics=['node_logged_in_users'],
logger=test_logger)
linter = Linter(test_logger)
profiles = linter.lint_profile_configs(test_profile_configs_path)
assert profiles['profile_0'] == expected_profile_0
assert profiles['profile_1'] == expected_profile_1
with pytest.raises(ValueError):
linter.lint_profile_configs(test_incorrect_profiles_path)
with pytest.raises(OSError):
linter.lint_profile_configs(test_invalid_path)
def run_garbage_collector(global_config_path: str, local_config_path: str, prom_dir_path: str, run_frequency: float,
internal_metrics_whitelist: Tuple[str], log_path: str, log_debug_path: str, ) -> None:
"""
Run the Garbage Collector
:param global_config_path: the path to the GlobalConfig object pulled down by the Config Service
:param local_config_path: the path to the local config object created by the Config Service
:param prom_dir_path: the path to the prom files directory where we look for stale prom files
:param run_frequency: the frequency to which we run the Garbage Collector
:param internal_metrics_whitelist: a list of internal metrics that we don't want the Garbage Collector to delete
:param log_path: the path to the scheduler log file
:param log_debug_path: the path to the iris.debug file that triggers when we want to enable verbose logging
:return: None
"""
logger = get_logger('iris.garbage_collector', log_path, log_debug_path)
prom_writer = PromFileWriter(logger=logger)
general_error_flag = False
while True:
try:
logger.info('Resuming the Garbage_Collector')
logger.info('Starting linter to transform the configs created by the config_service into python objs')
linter = Linter(logger=logger)
try:
global_config_obj = linter.lint_global_config(global_config_path)
local_config_obj = linter.lint_metrics_config(global_config_obj, local_config_path)
except OSError:
local_config_obj = {}
gc = GarbageCollector(
local_config_obj=local_config_obj,
internal_metrics_whitelist=internal_metrics_whitelist,
prom_dir_path=prom_dir_path,
logger=logger
)
logger.info('Running GC to detect stale prom files')
deleted_files = gc.delete_stale_prom_files()
metric_name = 'iris_garbage_collector_deleted_stale_files'
prom_builder = PromStrBuilder(
metric_name=metric_name,
metric_result=len(deleted_files),
help_str='Indicate how many stale prom files the Garbage Collector had to delete',
type_str='gauge'
)
prom_string = prom_builder.create_prom_string()
prom_file_path = os.path.join(prom_dir_path, '{}.prom'.format(metric_name))
prom_writer.write_prom_file(prom_file_path, prom_string)
general_error_flag = False
except Exception as e:
logger.error('Garbage Collector has an err: {}'.format(e))
general_error_flag = True
finally:
metric_name = 'iris_garbage_collector_error'
prom_builder = PromStrBuilder(
metric_name=metric_name,
metric_result=int(general_error_flag),
help_str='Indicate if an exception/error has occurred in the Garbage Collector',
type_str='gauge'
)
prom_string = prom_builder.create_prom_string()
prom_file_path = os.path.join(prom_dir_path, '{}.prom'.format(metric_name))
prom_writer.write_prom_file(prom_file_path, prom_string)
logger.info('Sleeping the Garbage Collector for {} seconds\n'.format(run_frequency))
time.sleep(run_frequency)
def run_config_service(aws_creds_path: str, s3_region_name: str,
s3_bucket_env: str, s3_bucket_name: str,
s3_download_to_path: str, ec2_region_name: str,
ec2_dev_instance_id: str, ec2_metadata_url: str,
local_config_path: str, prom_dir_path: str,
run_frequency: float, log_path: str,
log_debug_path: str, dev_mode: bool) -> None:
"""
Run the Config Service with each of its components (S3, Linter, EC2Tags)
:param aws_creds_path: path to the aws_credentials file
:param s3_region_name: region that the S3 bucket is in
:param s3_bucket_env: the bucket_environment/aws_profile_name, is the env the bucket is in ie prod/nonprod
:param s3_bucket_name: the name of the bucket
:param s3_download_to_path: the path to download the bucket content/configs
:param ec2_region_name: region that the ec2 instance is in
:param ec2_dev_instance_id: the instance id of the host you want to test in dev mode, see readme & iris.cfg. This
field is not set by default in iris.cfg when running on a host. It must be manually set by the tester for debugging
:param ec2_metadata_url: the metadata url that allows the instance to get info of itself, defined in iris.cfg
:param local_config_path: the path we want to write the local config object to. The local config object contains
the list of metrics the current ec2 host needs to run
:param prom_dir_path: the path to the prom files directory that we write metric results to
:param run_frequency: the frequency to which we run the config service
:param log_path: the path to the config_service log file
:param log_debug_path: the path to the iris.debug file that triggers when we want to enable verbose logging
:param dev_mode: set to True when you want to run in dev mode, see readme & iris.cfg
:return: None
"""
logger = get_logger('iris.config_service', log_path, log_debug_path)
general_error_flag = False
missing_iris_tags_error_flag = False
while True:
# Run config service S3 puller to get the config files from iris bucket
try:
logger.info('Resuming the Config_Service')
logger.info(
'Downloading content from s3 bucket: {} to dir: {}'.format(
s3_bucket_name, s3_download_to_path))
s3 = S3(aws_creds_path=aws_creds_path,
region_name=s3_region_name,
bucket_environment=s3_bucket_env,
bucket_name=s3_bucket_name,
dev_mode=dev_mode,
logger=logger)
s3.download_bucket(s3_download_to_path)
# run linter to transform downloaded s3 configs into Python objects. Also lints the configs for errors
global_config_path = os.path.join(s3_download_to_path,
'global_config.json')
metrics_config_path = os.path.join(s3_download_to_path,
'metrics.json')
profile_configs_path = os.path.join(s3_download_to_path,
'profiles')
logger.info(
'Starting linter to transform the downloaded configs into GlobalConfig, Metric, & Profile objs'
)
linter = Linter(logger)
logger.info(
'Linting Global Config file at {}'.format(global_config_path))
global_config = linter.lint_global_config(global_config_path)
logger.info('Linting Metrics Config file at {}'.format(
metrics_config_path))
metrics = linter.lint_metrics_config(global_config,
metrics_config_path)
logger.info('Linting Profile Configs file at {}'.format(
profile_configs_path))
profiles = linter.lint_profile_configs(profile_configs_path)
# run EC2Tags to retrieve the iris_tags of the host
logger.info('Retrieving current ec2 host iris_tags')
ec2 = EC2Tags(aws_creds_path=aws_creds_path,
region_name=ec2_region_name,
ec2_metadata_url=ec2_metadata_url,
dev_instance_id=ec2_dev_instance_id,
dev_mode=dev_mode,
logger=logger)
ec2_iris_tags = ec2.get_iris_tags()
# use iris_tags and downloaded s3 configs to generate the local_config object
logger.info(
'Matching retrieved iris_tags with the downloaded configs to generate the local_config obj'
)
iris_profile = ec2_iris_tags['ihr:iris:profile']
if iris_profile not in profiles:
err_msg = 'The ihr:iris:profile tag on {} is not defined in any profile config'.format(
ec2.instance_id)
logger.error(err_msg)
raise KeyError(err_msg)
local_config_metrics = {}
for prof_metric in profiles[iris_profile].metrics:
if prof_metric not in metrics:
err_msg = 'Metric {} in profile {} not defined in metrics config'.format(
prof_metric, iris_profile)
logger.error(err_msg)
raise KeyError(err_msg)
local_config_metrics[prof_metric] = metrics[
prof_metric].to_json()
logger.info('Generated the local_config object')
with tempfile.NamedTemporaryFile('w', delete=False) as tmpfile:
json.dump(local_config_metrics, tmpfile, indent=2)
os.rename(tmpfile.name, local_config_path)
logger.info('Finished writing to local_config file at {}'.format(
local_config_path))
general_error_flag = False
missing_iris_tags_error_flag = False
except MissingIrisTagsError as e:
logger.error('Config_Service MissingIrisTagsError: {}'.format(e))
missing_iris_tags_error_flag = True
# will log twice for defined err logs in iris, but will catch & log unlogged errs in code (3rd party err)
except Exception as e:
logger.error('Config_Service has an err: {}'.format(e))
general_error_flag = True
finally:
general_error_name = 'iris_config_service_error'
general_error_prom_builder = PromStrBuilder(
metric_name=general_error_name,
metric_result=int(general_error_flag),
help_str=
'Indicate if a general exception/error has occured in the Scheduler',
type_str='gauge')
general_error_prom_string = general_error_prom_builder.create_prom_string(
)
general_error_prom_file_path = os.path.join(
prom_dir_path, '{}.prom'.format(general_error_name))
missing_iris_tags_name = 'iris_missing_ec2_tags'
missing_iris_tags_prom_builder = PromStrBuilder(
metric_name=missing_iris_tags_name,
metric_result=int(missing_iris_tags_error_flag),
help_str='Indicate if the ec2 host is missing the iris tags',
type_str='gauge')
missing_iris_tags_prom_string = missing_iris_tags_prom_builder.create_prom_string(
)
missing_iris_tags_prom_file_path = os.path.join(
prom_dir_path, '{}.prom'.format(missing_iris_tags_name))
prom_writer = PromFileWriter(logger=logger)
prom_writer.write_prom_file(general_error_prom_file_path,
general_error_prom_string)
prom_writer.write_prom_file(missing_iris_tags_prom_file_path,
missing_iris_tags_prom_string)
logger.info(
'Sleeping the Config_Service for {}\n'.format(run_frequency))
time.sleep(run_frequency)
if __name__ == '__main__':
try:
logger.info('Running execute_metrics script for local development')
script_settings = util.read_config_file(os.path.join(project_root_dir, 'iris.cfg'), logger)
main_helpers.check_iris_test_settings(script_settings, logger)
iris_root_path = script_settings['main_settings']['iris_root_path']
local_config_path = os.path.join(iris_root_path, 'local_config.json')
global_config_path = os.path.join(os.path.join(iris_root_path, 'downloads'), 'global_config.json')
prom_dir_path = os.path.join(iris_root_path, 'prom_files')
logger.info('Starting linter to transform the config files pulled down from S3 into python objects')
linter = Linter(logger)
global_config_obj = linter.lint_global_config(global_config_path)
local_config_obj = linter.lint_metrics_config(global_config_obj, local_config_path)
metrics_list = list(local_config_obj.values())
logger.info('Executing metrics: {}'.format([metric.name for metric in metrics_list]))
scheduler = Scheduler(metrics_list, prom_dir_path, logger=logger)
scheduler.run()
except KeyError:
pass
finally:
logger.info('Finished running execute_metrics script')
def run_scheduler(global_config_path: str, local_config_path: str,
prom_dir_path: str, run_frequency: float,
internal_metrics_whitelist: Tuple[str], log_path: str,
log_debug_path: str) -> None:
"""
Run the Scheduler
:param global_config_path: the path to the GlobalConfig object pulled down by the Config Service
:param local_config_path: the path to the local config object created by the Config Service
:param prom_dir_path: the path to the prom files directory that we write metric results to
:param run_frequency: the frequency to which we run the scheduler
:param internal_metrics_whitelist: a list of internal metrics that we don't want to count in the
iris_custom_metrics_count metric
:param log_path: the path to the scheduler log file
:param log_debug_path: the path to the iris.debug file that triggers when we want to enable verbose logging
:return: None
"""
logger = get_logger('iris.scheduler', log_path, log_debug_path)
error_flag = 0
while True:
try:
sleep_total = 0 # the accumulated sleep time for checking the global_config and local_config
sleep_increment = 10 # check for global_config and local_config every 5 seconds if they don't exist
max_wait_time = 120 # max wait/sleep time that the scheduler will wait for these configs
while not os.path.isfile(global_config_path) or not os.path.isfile(
local_config_path):
if sleep_total == max_wait_time:
err_msg = 'No global_config: {} or local_config: {}. The scheduler has waited for 2 mins'.format(
global_config_path, local_config_path)
logger.error('OSError: {}'.format(err_msg))
raise OSError(err_msg)
else:
msg = 'The scheduler is still waiting on the config_service for the global_config/local_config file'
logger.warning(msg)
sleep_total += sleep_increment
time.sleep(sleep_increment)
# run linter to transform the local_config file created by the config_service into objects for the scheduler
logger.info(
'Starting linter to transform the config files created by the config_service into python objs'
)
linter = Linter(logger)
global_config_obj = linter.lint_global_config(global_config_path)
local_config_obj = linter.lint_metrics_config(
global_config_obj, local_config_path)
metrics_list = list(local_config_obj.values())
logger.info('Read local_config file metrics {}'.format(', '.join(
[metric.name for metric in metrics_list])))
# run scheduler to asynchronously execute each metric and asynchronously write to the metric's prom file
scheduler = Scheduler(metrics_list, prom_dir_path, logger=logger)
scheduler.run()
error_flag = 0
# will log twice for defined err logs in iris, but will catch & log unlogged errs in code (3rd party err)
except Exception as e:
logger.error('Scheduler has an err: {}'.format(e))
error_flag = 1
finally:
prom_writer = PromFileWriter(logger=logger)
metric_name = 'iris_scheduler_error'
prom_builder = PromStrBuilder(
metric_name=metric_name,
metric_result=error_flag,
help_str=
'Indicate if an exception/error has occured in the Scheduler',
type_str='gauge')
prom_string = prom_builder.create_prom_string()
prom_file_path = os.path.join(prom_dir_path,
'{}.prom'.format(metric_name))
prom_writer.write_prom_file(prom_file_path, prom_string)
# count how many custom metrics prom files are currently being exposed and create the prom file
custom_metrics_count_result = 0
for prom_file in os.listdir(prom_dir_path):
metric_name = prom_file.replace('.prom', '')
if metric_name not in internal_metrics_whitelist:
custom_metrics_count_result += 1
metric_name = 'iris_custom_metrics_count'
prom_builder = PromStrBuilder(
metric_name=metric_name,
metric_result=custom_metrics_count_result,
help_str=
'Indicate how many custom metrics the Scheduler is exposing',
type_str='gauge')
prom_string = prom_builder.create_prom_string()
prom_file_path = os.path.join(prom_dir_path,
'{}.prom'.format(metric_name))
prom_writer.write_prom_file(prom_file_path, prom_string)
logger.info('Sleeping the Scheduler for {} seconds\n'.format(
run_frequency))
time.sleep(run_frequency)