Airflow: Copy local file to GCS - airflow

I'm using Airflow to copy files from local to GCS, implemented using the FileToGoogleCloudStorageOperator
My code:
dag = DAG('file-gcs-demo', catchup=False, default_args=default_args)
cloud_storage_bucket_name = 'composerdemo-bucket'
t1 = FileToGoogleCloudStorageOperator(
task_id = 'copy-file-gcs',
src = '/Users/jw/data.json',
dst = '/data',
bucket = cloud_storage_bucket_name,
gcp_conn_id = 'google_cloud_default',
dag = dag
)
t1
The DAG runs successfully and when I check in GCS I can see a /data folder created in my bucket, but the folder is empty with no file copied. Any idea why this is happening?

dst is an object, not a folder. Change dst= to include an object name within the folder, for example dst="data/data.json".

Related

Apache Airflow unit and integration test

I am new to Apache Airflow and I am trying to figure out how to unit/integration test my dags/tasks
Here is my directory structure
/airflow
/dags
/tests/dags
I created a simple DAG which has a task to reads data from a Postgres table
def read_files(ti):
sql = "select id from files where status='NEW'"
pg_hook = PostgresHook(postgres_conn_id="metadata")
connection = pg_hook.get_conn()
cursor = connection.cursor()
cursor.execute(sql)
files = cursor.fetchall()
ti.xcom_push(key="files_to_process", value=files)
with DAG(dag_id="check_for_new_files", schedule_interval=timedelta(minutes=30),
start_date=datetime(2022, 9, 1), catchup=False) as dag:
check_files = PythonOperator(task_id="read_files",
python_callable=read_files)
Is it possible to test this by mocking Airflow/Postgres connection etc
yes it is possible to do test in dags, here is an example of basic things you can do:
import unittest
from airflow.models import DagBag
class TestCheckForNewFilesDAG(unittest.TestCase):
"""Check Dag"""
def setUp(self):
self.dagbag = DagBag()
def test_task_count(self):
"""Check task count for a dag"""
dag_id='check_for_new_files'
dag = self.dagbag.get_dag(dag_id)
self.assertEqual(len(dag.tasks), 1)
def test_contain_tasks(self):
"""Check task contains in hello_world dag"""
dag_id='check_for_new_files'
dag = self.dagbag.get_dag(dag_id)
tasks = dag.tasks
task_ids = list(map(lambda task: task.task_id, tasks))
self.assertListEqual(task_ids, ['read_files'])
def test_dependencies_of_read_files_task(self):
"""Check the task dependencies of a taskin hello_world dag"""
dag_id='check_for_new_files'
dag = self.dagbag.get_dag(dag_id)
read_files_task = dag.get_task('read_files')
# to be use in case you have upstream task
upstream_task_ids = list(map(lambda task: task.task_id,
read_files_task.upstream_list))
self.assertListEqual(upstream_task_ids, [])
downstream_task_ids = list(map(lambda task: task.task_id,
read_files_task.downstream_list))
self.assertListEqual(downstream_task_ids, [])
suite = unittest.TestLoader().loadTestsFromTestCase(TestHelloWorldDAG)
unittest.TextTestRunner(verbosity=2).run(suite)
In case of verifying that manipulated data of files are moved correctly the documentations suggest:
https://airflow.apache.org/docs/apache-airflow/2.0.1/best-practices.html#self-checks
Self-Checks
You can also implement checks in a DAG to make sure the tasks are producing the results as expected. As an example, if you have a task that pushes data to S3, you can implement a check in the next task. For example, the check could make sure that the partition is created in S3 and perform some simple checks to determine if the data is correct.
I think this is an excellent and straightforward way to verify a specific task.
Here there are other useful links you can use:
https://www.youtube.com/watch?v=ANJnYbLwLjE
In the next ones, they talk about mock
https://www.astronomer.io/guides/testing-airflow/
https://medium.com/#montadhar/apache-airflow-testing-guide-7956a3f4bbf5
https://godatadriven.com/blog/testing-and-debugging-apache-airflow/

How do I attach a file on email operator in Airflow

Whenever I add the files argument to the email_task I get a failed run.
email_task = EmailOperator(
task_id='email_sample_data',
to='sample#sample.com',
subject='Forecast for the day',
html_content= "Sample",
files=['/home/airflow/sample.html'],
dag=dag)
I'm getting an error that the file is not found. Where does airflow pick my file, where do I need to upload a file, and what is the correct syntax for the 'files' argument?
Airflow expect path to be relative to where the DAG file is stored.
However since files is templated field you can use template_search_path to provide additional paths that Airflow will look in:
with DAG(
...
template_searchpath = ['/home/airflow/'],
) as dag:
email_task = EmailOperator(
task_id='email_sample_data',
to='sample#sample.com',
subject='Forecast for the day',
html_content="Sample",
files=['/home/airflow/sample.html']
)

Problem Running MySQL Script with Airflow MySQL Operator

I'm not understanding something about using the MySQL operator to call and a MySQL script with Apache Airflow.
When I run this task...
MySQLTest = MySqlOperator(
task_id='MySQLTest',
sql = '/home/user/DBScripts/MySQLScript/SampleMySQLScript.sql',
mysql_conn_id = 'mysql_db_connect',
autocommit = True,
database = 'segments'
)
I get this error in traceback...
jinja2.exceptions.TemplateNotFound: /home/user/DBScripts/MySQLScript/SampleMySQLScript.sql
The DAG task runs fine if I provide the entire SQL script as a parameter.
I'm not familiar with Jinja templating.
Is it easier to learn to write my scripts as a Jinja template? Should I import the text of the script and assign it to a variable that I pass? Is there a way to write the Airflow task so that it isn't expecting a Jinja template?
This error message means that the .sql file is not found.
Using:
MySQLTest = MySqlOperator(
task_id='MySQLTest',
sql='test.sql',
mysql_conn_id='mysql_db_connect',
autocommit=True,
database='segments',
dag=dag
)
Where test.sql is located on the same folder as the DAG file works fine:
If the path of the .sql file is not relative to the DAG file you can use template_searchpath to define the list of folders (non relative) where jinja will look for the templates.
So Your code could look like:
default_args = { # pylint: disable=invalid-name
'owner': 'airflow',
'start_date': datetime(2020, 12, 03),
}
with DAG(
dag_id='my_sql_dag',
default_args=default_args,
schedule_interval=None,
template_searchpath=['/home/user/DBScripts/MySQLScript']
) as dag:
MySQLTest = MySqlOperator(
task_id='MySQLTest',
sql='SampleMySQLScript.sql',
mysql_conn_id='mysql_db_connect',
autocommit=True,
database='segments'
)

OCI (Oracle Cloud) Child Compartment

Question : Here is python code for oci (oracle cloud) . It is able to create Bucket or download and upload fine in bucket in root compartment . But i am not able to do the same on sub compartment. Sub compartment is " My_Sub_Compartment"
Please advise how to fix it .
import os
import oci
import io
from oci.config import from_file
data_dir = "D:\\DataScienceAndStats\\artificialintelligence\\CS223A"
files_to_process = [file for file in os.listdir(data_dir) if file.endswith('txt')]
bucket_name = "Sales_Data"
# this is to configure the oci configuration file
my_config = from_file(file_location="C:\\Users\\amits\\Desktop\\Oracle_Cloud\\config_file_oci.txt")
print(my_config)
# Test Configuration file of oci
# print(validate_config(my_config))
"""
Create object storage client and get its namespace
"""
object_storage_client = oci.object_storage.ObjectStorageClient(my_config)
namespace = object_storage_client.get_namespace().data
"""
Create a bucket if it does not exist
"""
try:
create_bucket_response = object_storage_client.create_bucket(namespace,
oci.object_storage.models.CreateBucketDetails(name=bucket_name, compartment_id=my_config['tenancy']))
except Exception as e:
print("Please read below messages")
print(e.message)
print(e.status)
"""
Uploading the files
"""
print("uploading files to bucket")
for upload_file in files_to_process:
print('Uploading file {}'.format(upload_file))
object_storage_client.put_object(namespace, bucket_name, upload_file, io.open(os.path.join(data_dir, upload_file),
'rb'))
"""
Listing a files in the Bucket
"""
object_list = object_storage_client.list_objects(namespace, bucket_name)
for o in object_list.data.objects:
print(o.name)
"""
Downloading files from Bucket
"""
object_name = "1.txt"
destination_dir = 'D:\\DataScienceAndStats\\artificialintelligence\\CS223A\\moved_files'.format(object_name)
get_obj = object_storage_client.get_object(namespace, bucket_name, object_name)
with open(os.path.join(destination_dir, object_name), 'wb') as f:
for chunk in get_obj.data.raw.stream(1024 * 1024, decode_content=False):
f.write(chunk)
Within create_bucket_response function you should provide the sub compartment OCID, instead of my_config['tenancy']. Have you tried this way? my_config['tenancy'] will always return the root compartment.
If you add compartment=ocid1.compartment.oc1..[...] (whatever OCID the target compartment has)
to your config_file_oci.txt and replace my_config['tenancy'] with my_config['compartment'] it should work.
You cannot have the same name for two buckets in the same namespace/tenancy, have you deleted the one you already created?

Removing Airflow task logs

I'm running 5 DAG's which have generated a total of about 6GB of log data in the base_log_folder over a months period. I just added a remote_base_log_folder but it seems it does not exclude logging to the base_log_folder.
Is there anyway to automatically remove old log files, rotate them or force airflow to not log on disk (base_log_folder) only in remote storage?
Please refer https://github.com/teamclairvoyant/airflow-maintenance-dags
This plugin has DAGs that can kill halted tasks and log-cleanups.
You can grab the concepts and can come up with a new DAG that can cleanup as per your requirement.
We remove the Task logs by implementing our own FileTaskHandler, and then pointing to it in the airflow.cfg. So, we overwrite the default LogHandler to keep only N task logs, without scheduling additional DAGs.
We are using Airflow==1.10.1.
[core]
logging_config_class = log_config.LOGGING_CONFIG
log_config.LOGGING_CONFIG
BASE_LOG_FOLDER = conf.get('core', 'BASE_LOG_FOLDER')
FOLDER_TASK_TEMPLATE = '{{ ti.dag_id }}/{{ ti.task_id }}'
FILENAME_TEMPLATE = '{{ ti.dag_id }}/{{ ti.task_id }}/{{ ts }}/{{ try_number }}.log'
LOGGING_CONFIG = {
'formatters': {},
'handlers': {
'...': {},
'task': {
'class': 'file_task_handler.FileTaskRotationHandler',
'formatter': 'airflow.job',
'base_log_folder': os.path.expanduser(BASE_LOG_FOLDER),
'filename_template': FILENAME_TEMPLATE,
'folder_task_template': FOLDER_TASK_TEMPLATE,
'retention': 20
},
'...': {}
},
'loggers': {
'airflow.task': {
'handlers': ['task'],
'level': JOB_LOG_LEVEL,
'propagate': False,
},
'airflow.task_runner': {
'handlers': ['task'],
'level': LOG_LEVEL,
'propagate': True,
},
'...': {}
}
}
file_task_handler.FileTaskRotationHandler
import os
import shutil
from airflow.utils.helpers import parse_template_string
from airflow.utils.log.file_task_handler import FileTaskHandler
class FileTaskRotationHandler(FileTaskHandler):
def __init__(self, base_log_folder, filename_template, folder_task_template, retention):
"""
:param base_log_folder: Base log folder to place logs.
:param filename_template: template filename string.
:param folder_task_template: template folder task path.
:param retention: Number of folder logs to keep
"""
super(FileTaskRotationHandler, self).__init__(base_log_folder, filename_template)
self.retention = retention
self.folder_task_template, self.folder_task_template_jinja_template = \
parse_template_string(folder_task_template)
#staticmethod
def _get_directories(path='.'):
return next(os.walk(path))[1]
def _render_folder_task_path(self, ti):
if self.folder_task_template_jinja_template:
jinja_context = ti.get_template_context()
return self.folder_task_template_jinja_template.render(**jinja_context)
return self.folder_task_template.format(dag_id=ti.dag_id, task_id=ti.task_id)
def _init_file(self, ti):
relative_path = self._render_folder_task_path(ti)
folder_task_path = os.path.join(self.local_base, relative_path)
subfolders = self._get_directories(folder_task_path)
to_remove = set(subfolders) - set(subfolders[-self.retention:])
for dir_to_remove in to_remove:
full_dir_to_remove = os.path.join(folder_task_path, dir_to_remove)
print('Removing', full_dir_to_remove)
shutil.rmtree(full_dir_to_remove)
return FileTaskHandler._init_file(self, ti)
Airflow maintainers don't think truncating logs is a part of airflow core logic, to see this, and then in this issue, maintainers suggest to change LOG_LEVEL avoid too many log data.
And in this PR, we can learn how to change log level in airflow.cfg.
good luck.
I know it sounds savage, but have you tried pointing base_log_folder to /dev/null? I use Airflow as a part of a container, so I don't care about the files either, as long as the logger pipe to STDOUT as well.
Not sure how well this plays with S3 though.
For your concrete problems, I have some suggestions.
For those, you would always need a specialized logging config as described in this answer: https://stackoverflow.com/a/54195537/2668430
automatically remove old log files and rotate them
I don't have any practical experience with the TimedRotatingFileHandler from the Python standard library yet, but you might give it a try:
https://docs.python.org/3/library/logging.handlers.html#timedrotatingfilehandler
It not only offers to rotate your files based on a time interval, but if you specify the backupCount parameter, it even deletes your old log files:
If backupCount is nonzero, at most backupCount files will be kept, and if more would be created when rollover occurs, the oldest one is deleted. The deletion logic uses the interval to determine which files to delete, so changing the interval may leave old files lying around.
Which sounds pretty much like the best solution for your first problem.
force airflow to not log on disk (base_log_folder), but only in remote storage?
In this case you should specify the logging config in such a way that you do not have any logging handlers that write to a file, i.e. remove all FileHandlers.
Rather, try to find logging handlers that send the output directly to a remote address.
E.g. CMRESHandler which logs directly to ElasticSearch but needs some extra fields in the log calls.
Alternatively, write your own handler class and let it inherit from the Python standard library's HTTPHandler.
A final suggestion would be to combine both the TimedRotatingFileHandler and setup ElasticSearch together with FileBeat, so you would be able to store your logs inside ElasticSearch (i.e. remote), but you wouldn't store a huge amount of logs on your Airflow disk since they will be removed by the backupCount retention policy of your TimedRotatingFileHandler.
Usually apache airflow grab the disk space due to 3 reasons
1. airflow scheduler logs files
2. mysql binaly logs [Major]
3. xcom table records.
To make it clean up on regular basis I have set up a dag which run on daily basis and cleans the binary logs and truncate the xcom table to make the disk space free
You also might need to install [pip install mysql-connector-python].
To clean up scheduler log files I do delete them manually two times in a week to avoid the risk of logs deleted which needs to be required for some reasons.
I clean the logs files by [sudo rm -rd airflow/logs/] command.
Below is my python code for reference
'
"""Example DAG demonstrating the usage of the PythonOperator."""
from airflow import DAG
from airflow.operators.python import PythonOperator
from datetime import datetime, timedelta
from airflow.utils.dates import days_ago
from airflow.operators.bash import BashOperator
from airflow.providers.postgres.operators.postgres import PostgresOperator
args = {
'owner': 'airflow',
'email_on_failure':True,
'retries': 1,
'email':['Your Email Id'],
'retry_delay': timedelta(minutes=5)
}
dag = DAG(
dag_id='airflow_logs_cleanup',
default_args=args,
schedule_interval='#daily',
start_date=days_ago(0),
catchup=False,
max_active_runs=1,
tags=['airflow_maintenance'],
)
def truncate_table():
import mysql.connector
connection = mysql.connector.connect(host='localhost',
database='db_name',
user='username',
password='your password',
auth_plugin='mysql_native_password')
cursor = connection.cursor()
sql_select_query = """TRUNCATE TABLE xcom"""
cursor.execute(sql_select_query)
connection.commit()
connection.close()
print("XCOM Table truncated successfully")
def delete_binary_logs():
import mysql.connector
from datetime import datetime
date = datetime.today().strftime('%Y-%m-%d')
connection = mysql.connector.connect(host='localhost',
database='db_name',
user='username',
password='your_password',
auth_plugin='mysql_native_password')
cursor = connection.cursor()
query = 'PURGE BINARY LOGS BEFORE ' + "'" + str(date) + "'"
sql_select_query = query
cursor.execute(sql_select_query)
connection.commit()
connection.close()
print("Binary logs deleted successfully")
t1 = PythonOperator(
task_id='truncate_table',
python_callable=truncate_table, dag=dag
)
t2 = PythonOperator(
task_id='delete_binary_logs',
python_callable=delete_binary_logs, dag=dag
)
t2 << t1
'
I am surprized but it worked for me. Update your config as below:
base_log_folder=""
It is test in minio and in s3.
Our solution looks a lot like Franzi's:
Running on Airflow 2.0.1 (py3.8)
Override default logging configuration
Since we use a helm chart for airflow deployment it was easiest to push an env there, but it can also be done in the airflow.cfg or using ENV in dockerfile.
# Set custom logging configuration to enable log rotation for task logging
AIRFLOW__LOGGING__LOGGING_CONFIG_CLASS: "airflow_plugins.settings.airflow_local_settings.DEFAULT_LOGGING_CONFIG"
Then we added the logging configuration together with the custom log handler to a python module we build and install in the docker image. As described here: https://airflow.apache.org/docs/apache-airflow/stable/modules_management.html
Logging configuration snippet
This is only a copy on the default from the airflow codebase, but then the task logger gets a different handler.
DEFAULT_LOGGING_CONFIG: Dict[str, Any] = {
'version': 1,
'disable_existing_loggers': False,
'formatters': {
'airflow': {'format': LOG_FORMAT},
'airflow_coloured': {
'format': COLORED_LOG_FORMAT if COLORED_LOG else LOG_FORMAT,
'class': COLORED_FORMATTER_CLASS if COLORED_LOG else 'logging.Formatter',
},
},
'handlers': {
'console': {
'class': 'airflow.utils.log.logging_mixin.RedirectStdHandler',
'formatter': 'airflow_coloured',
'stream': 'sys.stdout',
},
'task': {
'class': 'airflow_plugins.log.rotating_file_task_handler.RotatingFileTaskHandler',
'formatter': 'airflow',
'base_log_folder': os.path.expanduser(BASE_LOG_FOLDER),
'filename_template': FILENAME_TEMPLATE,
'maxBytes': 10485760, # 10MB
'backupCount': 6,
},
...
RotatingFileTaskHandler
And finally the custom handler which is just a merge of the logging.handlers.RotatingFileHandler and the FileTaskHandler.
#
# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements. See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership. The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied. See the License for the
# specific language governing permissions and limitations
# under the License.
"""File logging handler for tasks."""
import logging
import os
from pathlib import Path
from typing import TYPE_CHECKING, Optional
import requests
from airflow.configuration import AirflowConfigException, conf
from airflow.utils.helpers import parse_template_string
if TYPE_CHECKING:
from airflow.models import TaskInstance
class RotatingFileTaskHandler(logging.Handler):
"""
FileTaskHandler is a python log handler that handles and reads
task instance logs. It creates and delegates log handling
to `logging.FileHandler` after receiving task instance context.
It reads logs from task instance's host machine.
:param base_log_folder: Base log folder to place logs.
:param filename_template: template filename string
"""
def __init__(self, base_log_folder: str, filename_template: str, maxBytes=0, backupCount=0):
self.max_bytes = maxBytes
self.backup_count = backupCount
super().__init__()
self.handler = None # type: Optional[logging.FileHandler]
self.local_base = base_log_folder
self.filename_template, self.filename_jinja_template = parse_template_string(filename_template)
def set_context(self, ti: "TaskInstance"):
"""
Provide task_instance context to airflow task handler.
:param ti: task instance object
"""
local_loc = self._init_file(ti)
self.handler = logging.handlers.RotatingFileHandler(
filename=local_loc,
mode='a',
maxBytes=self.max_bytes,
backupCount=self.backup_count,
encoding='utf-8',
delay=False,
)
if self.formatter:
self.handler.setFormatter(self.formatter)
self.handler.setLevel(self.level)
def emit(self, record):
if self.handler:
self.handler.emit(record)
def flush(self):
if self.handler:
self.handler.flush()
def close(self):
if self.handler:
self.handler.close()
def _render_filename(self, ti, try_number):
if self.filename_jinja_template:
if hasattr(ti, 'task'):
jinja_context = ti.get_template_context()
jinja_context['try_number'] = try_number
else:
jinja_context = {
'ti': ti,
'ts': ti.execution_date.isoformat(),
'try_number': try_number,
}
return self.filename_jinja_template.render(**jinja_context)
return self.filename_template.format(
dag_id=ti.dag_id,
task_id=ti.task_id,
execution_date=ti.execution_date.isoformat(),
try_number=try_number,
)
def _read_grouped_logs(self):
return False
def _read(self, ti, try_number, metadata=None): # pylint: disable=unused-argument
"""
Template method that contains custom logic of reading
logs given the try_number.
:param ti: task instance record
:param try_number: current try_number to read log from
:param metadata: log metadata,
can be used for steaming log reading and auto-tailing.
:return: log message as a string and metadata.
"""
# Task instance here might be different from task instance when
# initializing the handler. Thus explicitly getting log location
# is needed to get correct log path.
log_relative_path = self._render_filename(ti, try_number)
location = os.path.join(self.local_base, log_relative_path)
log = ""
if os.path.exists(location):
try:
with open(location) as file:
log += f"*** Reading local file: {location}\n"
log += "".join(file.readlines())
except Exception as e: # pylint: disable=broad-except
log = f"*** Failed to load local log file: {location}\n"
log += "*** {}\n".format(str(e))
elif conf.get('core', 'executor') == 'KubernetesExecutor': # pylint: disable=too-many-nested-blocks
try:
from airflow.kubernetes.kube_client import get_kube_client
kube_client = get_kube_client()
if len(ti.hostname) >= 63:
# Kubernetes takes the pod name and truncates it for the hostname. This truncated hostname
# is returned for the fqdn to comply with the 63 character limit imposed by DNS standards
# on any label of a FQDN.
pod_list = kube_client.list_namespaced_pod(conf.get('kubernetes', 'namespace'))
matches = [
pod.metadata.name
for pod in pod_list.items
if pod.metadata.name.startswith(ti.hostname)
]
if len(matches) == 1:
if len(matches[0]) > len(ti.hostname):
ti.hostname = matches[0]
log += '*** Trying to get logs (last 100 lines) from worker pod {} ***\n\n'.format(
ti.hostname
)
res = kube_client.read_namespaced_pod_log(
name=ti.hostname,
namespace=conf.get('kubernetes', 'namespace'),
container='base',
follow=False,
tail_lines=100,
_preload_content=False,
)
for line in res:
log += line.decode()
except Exception as f: # pylint: disable=broad-except
log += '*** Unable to fetch logs from worker pod {} ***\n{}\n\n'.format(ti.hostname, str(f))
else:
url = os.path.join("http://{ti.hostname}:{worker_log_server_port}/log", log_relative_path).format(
ti=ti, worker_log_server_port=conf.get('celery', 'WORKER_LOG_SERVER_PORT')
)
log += f"*** Log file does not exist: {location}\n"
log += f"*** Fetching from: {url}\n"
try:
timeout = None # No timeout
try:
timeout = conf.getint('webserver', 'log_fetch_timeout_sec')
except (AirflowConfigException, ValueError):
pass
response = requests.get(url, timeout=timeout)
response.encoding = "utf-8"
# Check if the resource was properly fetched
response.raise_for_status()
log += '\n' + response.text
except Exception as e: # pylint: disable=broad-except
log += "*** Failed to fetch log file from worker. {}\n".format(str(e))
return log, {'end_of_log': True}
def read(self, task_instance, try_number=None, metadata=None):
"""
Read logs of given task instance from local machine.
:param task_instance: task instance object
:param try_number: task instance try_number to read logs from. If None
it returns all logs separated by try_number
:param metadata: log metadata,
can be used for steaming log reading and auto-tailing.
:return: a list of listed tuples which order log string by host
"""
# Task instance increments its try number when it starts to run.
# So the log for a particular task try will only show up when
# try number gets incremented in DB, i.e logs produced the time
# after cli run and before try_number + 1 in DB will not be displayed.
if try_number is None:
next_try = task_instance.next_try_number
try_numbers = list(range(1, next_try))
elif try_number < 1:
logs = [
[('default_host', f'Error fetching the logs. Try number {try_number} is invalid.')],
]
return logs, [{'end_of_log': True}]
else:
try_numbers = [try_number]
logs = [''] * len(try_numbers)
metadata_array = [{}] * len(try_numbers)
for i, try_number_element in enumerate(try_numbers):
log, metadata = self._read(task_instance, try_number_element, metadata)
# es_task_handler return logs grouped by host. wrap other handler returning log string
# with default/ empty host so that UI can render the response in the same way
logs[i] = log if self._read_grouped_logs() else [(task_instance.hostname, log)]
metadata_array[i] = metadata
return logs, metadata_array
def _init_file(self, ti):
"""
Create log directory and give it correct permissions.
:param ti: task instance object
:return: relative log path of the given task instance
"""
# To handle log writing when tasks are impersonated, the log files need to
# be writable by the user that runs the Airflow command and the user
# that is impersonated. This is mainly to handle corner cases with the
# SubDagOperator. When the SubDagOperator is run, all of the operators
# run under the impersonated user and create appropriate log files
# as the impersonated user. However, if the user manually runs tasks
# of the SubDagOperator through the UI, then the log files are created
# by the user that runs the Airflow command. For example, the Airflow
# run command may be run by the `airflow_sudoable` user, but the Airflow
# tasks may be run by the `airflow` user. If the log files are not
# writable by both users, then it's possible that re-running a task
# via the UI (or vice versa) results in a permission error as the task
# tries to write to a log file created by the other user.
relative_path = self._render_filename(ti, ti.try_number)
full_path = os.path.join(self.local_base, relative_path)
directory = os.path.dirname(full_path)
# Create the log file and give it group writable permissions
# TODO(aoen): Make log dirs and logs globally readable for now since the SubDag
# operator is not compatible with impersonation (e.g. if a Celery executor is used
# for a SubDag operator and the SubDag operator has a different owner than the
# parent DAG)
Path(directory).mkdir(mode=0o777, parents=True, exist_ok=True)
if not os.path.exists(full_path):
open(full_path, "a").close()
# TODO: Investigate using 444 instead of 666.
os.chmod(full_path, 0o666)
return full_path
Maybe a final note; the links in the airflow UI to the logging will now only open the latest logfile, not the older rotated files which are only accessible by means of SSH or any other interface to access the airflow logging path.
I don't think that there is a rotation mechanism but you can store them in S3 or google cloud storage as describe here : https://airflow.incubator.apache.org/configuration.html#logs

Resources