Praetor is a minimal Web UI for monitoring Prefect Core flows. Airflow is so awesome because of it's UI and it's sad that even more awesome Prefect Core lacks one.
This is strictly for monitoring task states and looking at logs. If you need a fully managed solution, consider using Prefect Cloud.
There are two parts to Praetor: webserver and custom classes. Webserver lets you look at what's going on with your flows and runner is used to run flows. Custom classes are the same as normal Prefect classes: Flow, Task and @task decorator. The difference is that they report their state to the webserver.
I tried to improve on the parts of Airflow UI that I didn't like:
- It's realtime and updates itself, so no need to manually hit refresh every 5 seconds.
- In tree view (it's the only option for now) one task is displayed exactly once. Tasks are topologically sorted, so all displayed connections are top-down.
- Handle your flows for you (starting, restarts etc.)
- Manage your secrets
- Authenticate users
To start webserver, run
$ praetor webserver --dask localhost:8786--dask is optional. If you don't provide the dask option, the server will only be available in local mode.
To run a flow, do:
from praetor import Flow, Task, task # instead of: from prefect import Flow, Task, taskthen proceed as usual.
These are subclasses of normal prefect classes, but with additional state handlers and hooks for reporting start/stop of the flow. If you use DaskExecutor, the flow will push update messages into a Queue on the dask cluster. If you're not using it, it will try and use a simple REST API at http://localhost:8000, which is called "local mode".
For now, clone and run $ python setup.py install -e '.[webserver]. Note that this is not ready for production at all, so use at your own peril.
Note that both prefect and praetor need to be installed on Dask workers. See examples folder for an example Dockerfile and Compose setup.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent