A just another Cron alternative with a Web UI, but with much more capabilities
It runs DAGs (Directed acyclic graph) defined in a simple YAML format.
- Install by placing just a single binary file
- Schedule executions of DAGs with Cron expressions
- Define dependencies between related jobs and represent them as a single DAG (unit of execution)
- Highlights
- Contents
- Getting started
- Motivation
- Why not existing workflow schedulers, such as Airflow?
- How does it work?
- Install
dagu
- ️Quick start
- Command Line User Interface
- Web User Interface
- YAML format
- Admin Configuration
- Environment Variable
- Base Configuration for all DAGs
- Scheduler
- REST API Interface
- FAQ
- How to contribute?
- Where is the history data stored?
- Where are the log files stored?
- How long will the history data be stored?
- How to use specific
host
andport
fordagu server
? - How to specify the DAGs directory for
dagu server
anddagu scheduler
? - How can I retry a DAG from a specific task?
- How does it track running processes without DBMS?
- License
- Contributors
See Install dagu
and ️Quick start.
In the projects I worked on, our ETL pipeline had many problems. There were hundreds of cron jobs on the server's crontab, and it is impossible to keep track of those dependencies between them. If one job failed, we were not sure which to rerun. We also have to SSH into the server to see the logs and run each shell script one by one. So we needed a tool that could explicitly visualize and manage the dependencies of the pipeline. How nice it would be to be able to visually see the job dependencies, execution status, and logs of each job in a Web UI and to be able to rerun or stop a series of jobs with just a mouse click!
There are existing tools such as Airflow, Prefect, Temporal, etc, but in most cases they require writing code in a programming language such as Python to define DAGs. In systems that have been in operation for a long time, there are already complex jobs written in hundreds of thousands of lines of code in other languages such as Perl or Shell Scripts, and there is concern that adding another layer of Python code will further decrease maintainability. So we developed Dagu, which requires no coding, and is easy-to-use and self-contained, making it ideal for smaller projects with fewer people.
Dagu is a single command and it uses the local file system to store data. Therefore, no DBMS or cloud service is required. Dagu executes DAGs defined in declarative YAML format. Existing programs can be used without any modification.
You can quickly install dagu
command and try it out.
brew install yohamta/tap/dagu
Upgrade to the latest version:
brew upgrade yohamta/tap/dagu
curl -L https://raw.githubusercontent.com/yohamta/dagu/main/scripts/downloader.sh | bash
Download the latest binary from the Releases page and place it in your $PATH
(e.g. /usr/local/bin
).
Start the server with dagu server
and browse to http://127.0.0.1:8080
to explore the Web UI.
Create a DAG by clicking the New DAG
button on the top page of the web UI. Input example
in the dialog.
Note: DAG (YAML) files will be placed in ~/.dagu/dags
by default. See Admin Configuration for more details.
Go to the SPEC
Tab and hit the Edit
button. Copy & Paste this example YAML and click the Save
button.
You can execute the example by pressing the Start
button
dagu start [--params=<params>] <file>
- Runs the DAGdagu status <file>
- Displays the current status of the DAGdagu retry --req=<request-id> <file>
- Re-runs the specified DAG rundagu stop <file>
- Stops the DAG execution by sending TERM signalsdagu dry [--params=<params>] <file>
- Dry-runs the DAGdagu server [--host=<host>] [--port=<port>] [--dags=<path/to/the DAGs directory>]
- Starts the web server for web UIdagu scheduler [--dags=<path/to/the DAGs directory>]
- Starts the scheduler processdagu version
- Shows the current binary version
The --config=<config>
option is available to all commands. It allows to specify different Dagu configuration for the commands. Which enables you to manage multiple Dagu process in a single instance. See Admin Configuration for more details.
For example:
dagu server --config=~/.dagu/dev.yaml
dagu scheduler --config=~/.dagu/dev.yaml
-
DAGs: It shows all DAGs and the real-time status.
-
DAG Details: It shows the real-time status, logs, and DAG configurations. You can edit DAG configurations on a browser.
You can switch to the vertical graph with the button on the top right corner.
-
Search DAGs: It greps given text across all DAGs definitions.
-
Execution History: It shows past execution results and logs.
-
DAG Execution Log: It shows the detail log and standard output of each execution and step.
The minimal DAG definition is as simple as follows:
steps:
- name: step 1
command: echo hello
- name: step 2
command: echo world
depends:
- step 1
script
field provides a way to run arbitrary snippets of code in any language.
steps:
- name: step 1
command: "bash"
script: |
cd /tmp
echo "hello world" > hello
cat hello
output: RESULT
- name: step 2
command: echo ${RESULT} # hello world
depends:
- step 1
You can define environment variables and refer to using env
field.
env:
- SOME_DIR: ${HOME}/batch
- SOME_FILE: ${SOME_DIR}/some_file
steps:
- name: some task in some dir
dir: ${SOME_DIR}
command: python main.py ${SOME_FILE}
You can define parameters using params
field and refer to each parameter as $1, $2, etc. Parameters can also be command substitutions or environment variables. It can be overridden by --params=
parameter of start
command.
params: param1 param2
steps:
- name: some task with parameters
command: python main.py $1 $2
Named parameters are also available as follows:
params: ONE=1 TWO=`echo 2`
steps:
- name: some task with parameters
command: python main.py $ONE $TWO
You can use command substitution in field values. I.e., a string enclosed in backquotes (`
) is evaluated as a command and replaced with the result of standard output.
env:
TODAY: "`date '+%Y%m%d'`"
steps:
- name: hello
command: "echo hello, today is ${TODAY}"
Sometimes you have parts of a DAG that you only want to run under certain conditions. You can use the precondition
field to add conditional branches to your DAG.
For example, the below task only runs on the first date of each month.
steps:
- name: A monthly task
command: monthly.sh
preconditions:
- condition: "`date '+%d'`"
expected: "01"
If you want the DAG to continue to the next step regardless of the step's conditional check result, you can use the continueOn
field:
steps:
- name: A monthly task
command: monthly.sh
preconditions:
- condition: "`date '+%d'`"
expected: "01"
continueOn:
skipped: true
output
field can be used to set a environment variable with standard output. Leading and trailing space will be trimmed automatically. The environment variables can be used in subsequent steps.
steps:
- name: step 1
command: "echo foo"
output: FOO # will contain "foo"
stdout
field can be used to write standard output to a file.
steps:
- name: create a file
command: "echo hello"
stdout: "/tmp/hello" # the content will be "hello\n"
It is often desirable to take action when a specific event happens, for example, when a DAG fails. To achieve this, you can use handlerOn
fields.
handlerOn:
failure:
command: notify_error.sh
exit:
command: cleanup.sh
steps:
- name: A task
command: main.sh
If you want a task to repeat execution at regular intervals, you can use the repeatPolicy
field. If you want to stop the repeating task, you can use the stop
command to gracefully stop the task.
steps:
- name: A task
command: main.sh
repeatPolicy:
repeat: true
intervalSec: 60
You can call other DAGs in the same directory by using dagu start
command (you can omit .yaml
).
steps:
- name: Sub DAG
command: dagu start other_dag
If you want to call DAGs in other directory you can specify the DAG by absolute path.
steps:
- name: Sub DAG
command: dagu start /path/to/dag.yaml
Combining these settings gives you granular control over how the DAG runs.
name: all configuration # Name (optional, default is filename)
description: run a DAG # Description
schedule: "0 * * * *" # Execution schedule (cron expression)
group: DailyJobs # Group name to organize DAGs (optional)
tags: example # Free tags (separated by comma)
env: # Environment variables
- LOG_DIR: ${HOME}/logs
- PATH: /usr/local/bin:${PATH}
logDir: ${LOG_DIR} # Log directory to write standard output, default: ${DAG_HOME}/logs/dags
histRetentionDays: 3 # Execution history retention days (not for log files)
delaySec: 1 # Interval seconds between steps
maxActiveRuns: 1 # Max parallel number of running step
params: param1 param2 # Default parameters that can be referred to by $1, $2, ...
preconditions: # Precondisions for whether the it is allowed to run
- condition: "`echo $2`" # Command or variables to evaluate
expected: "param2" # Expected value for the condition
mailOn:
failure: true # Send a mail when the it failed
success: true # Send a mail when the it finished
MaxCleanUpTimeSec: 300 # The maximum amount of time to wait after sending a TERM signal to running steps before killing them
handlerOn: # Handlers on Success, Failure, Cancel, and Exit
success:
command: "echo succeed" # Command to execute when the execution succeed
failure:
command: "echo failed" # Command to execute when the execution failed
cancel:
command: "echo canceled" # Command to execute when the execution canceled
exit:
command: "echo finished" # Command to execute when the execution finished
steps:
- name: some task # Step name
description: some task # Step description
dir: ${HOME}/logs # Working directory (default: the same directory of the DAG file)
command: bash # Command and parameters
stdout: /tmp/outfile
ouptut: RESULT_VARIABLE
script: |
echo "any script"
signalOnStop: "SIGINT" # Specify signal name (e.g. SIGINT) to be sent when process is stopped
mailOn:
failure: true # Send a mail when the step failed
success: true # Send a mail when the step finished
continueOn:
failure: true # Continue to the next regardless of the step failed or not
skipped: true # Continue to the next regardless the preconditions are met or not
retryPolicy: # Retry policy for the step
limit: 2 # Retry up to 2 times when the step failed
intervalSec: 5 # Interval time before retry
repeatPolicy: # Repeat policy for the step
repeat: true # Boolean whether to repeat this step
intervalSec: 60 # Interval time to repeat the step in seconds
preconditions: # Precondisions for whether the step is allowed to run
- condition: "`echo $1`" # Command or variables to evaluate
expected: "param1" # Expected Value for the condition
The global configuration file ~/.dagu/config.yaml
is useful to gather common settings, such as logDir
or env
.
To configure Dagu, please create the config file (default path: ~/.dagu/admin.yaml
). All fields are optional.
# Web Server Host and Port
host: <hostname for web UI address> # default: 127.0.0.1
port: <port number for web UI address> # default: 8000
# path to the DAGs directory
dags: <the location of DAG configuration files> # default: ${DAG_HOME}/dags
# Web UI Color & Title
navbarColor: <admin-web header color> # header color for web UI (e.g. "#ff0000")
navbarTitle: <admin-web title text> # header title for web UI (e.g. "PROD")
# Basic Auth
isBasicAuth: <true|false> # enables basic auth
basicAuthUsername: <username for basic auth of web UI> # basic auth user
basicAuthPassword: <password for basic auth of web UI> # basic auth password
# Base Config
baseConfig: <base DAG config path> . # default: ${DAG_HOME}/config.yaml
# Others
logDir: <internal logdirectory> # default: ${DAG_HOME}/logs/admin
command: <Absolute path to the dagu binary> # default: dagu
You can configure the Dagu's internal work directory by defining DAGU_HOME
environment variables. Default path is ~/.dagu/
.
Creating a base configuration (default path: ~/.dagu/config.yaml
) is a convenient way to organize shared settings among all DAGs. The path to the base configuration file can be configured. See Admin Configuration for more details.
logDir: <path-to-write-log> # log directory to write standard output
histRetentionDays: 3 # history retention days
smtp: # [optional] mail server configuration to send notifications
host: <smtp server host>
port: <stmp server port>
errorMail: # [optional] mail configuration for error-level
from: <from address>
to: <to address>
prefix: <prefix of mail subject>
infoMail:
from: <from address> # [optional] mail configuration for info-level
to: <to address>
prefix: <prefix of mail subject>
To run DAGs automatically, you need to run dagu scheduler
process on your system.
You can specify the schedule with cron expression in the schedule
field in the config file as follows:
schedule: "5 4 * * *" # Run at 04:05.
steps:
- name: scheduled job
command: job.sh
Or you can set multiple schedules:
schedule:
- "30 7 * * *" # Run at 7:30
- "0 20 * * *" # Also run at 20:00
steps:
- name: scheduled job
command: job.sh
The easiest way to make sure the process is always running on your system is to create the script below and execute it every minute using cron (you don't need root
account in this way):
#!/bin/bash
process="dagu scheduler"
command="/usr/bin/dagu scheduler"
if ps ax | grep -v grep | grep "$process" > /dev/null
then
exit
else
$command &
fi
exit
Set the dags
field to specify the directory of the DAGs.
dags: <the location of DAG configuration files> # default: (~/.dagu/dags)
Please refer to REST API Docs
Feel free to contribute in any way you want. Share ideas, questions, submit issues, and create pull requests. Thanks!
It will store execution history data in the DAGU__DATA
environment variable path. The default location is $HOME/.dagu/data
.
It will store log files in the DAGU__LOGS
environment variable path. The default location is $HOME/.dagu/logs
. You can override the setting by the logDir
field in a YAML file.
The default retention period for execution history is 30 days. However, you can override the setting by the histRetentionDays
field in a YAML file.
dagu server's host and port can be configured in the admin configuration file as below. See Admin Configuration for more details.
host: <hostname for web UI address> # default: 127.0.0.1
port: <port number for web UI address> # default: 8000
You can customize DAGs directory that will be used by dagu server
and dagu scheduler
. See Admin Configuration for more details.
dags: <the location of DAG configuration files> # default: ${DAG_HOME}/dags
You can change the status of any task to a failed
state. Then, when you retry the DAG, it will execute the failed one and any subsequent.
Dagu uses Unix sockets to communicate with running processes.
This project is licensed under the GNU GPLv3 - see the LICENSE.md file for details
Made with contrib.rocks.