I've submitted a PR at https://github.com/apache/airflow/pull/19843.

The code was written quite a while ago. Airflow 2.2 has since been released
but I've not noticed any regressions on rebasing over it.

Would appreciate any reviews. Thanks!

On Sat, 28 Aug 2021 at 23:19, Chris Fei <[email protected]> wrote:

> I do something similar to what you describe, where I can replace what's
> displayed in the "Code" view based on some other content. I was able to
> cobble it together by using airflow_local_settings.py to patch the function
> that's invoked when the "Code" view is accessed. It's a total hack, but it
> works for my use-case (I'm on airflow 2.1.3). YMMV—
>
> I have a python file in the dags folder that parses YAML files from a
> directory and publishes DAGs to the global scope based on the content of
> each YAML file (based on your previous messages, you probably have
> something similar). When creating each DAG, I add a KV pair in default_args
> for the file path to the underlying YAML. To access the YAML file in the
> Code view on the webserver, I set up airflow_local_settings.py to monkey
> patch the airflow.www.views.Airflow.code function with my own version. The
> patched function sets up a DagBag to allows me to access the full DAG
> object, where I can then read the YAML file from default_args and then
> return its content in the output. In my case, the scheduler and webserver
> run on the same VM so I can pass just the file path between them. If you
> can't just pass around the file path, I suppose you could try tossing the
> entire YAML content as a default_arg as well.
>
> This approach doesn't change fileloc and doesn't output the YAML file
> content to the dag_code table.
>
> One obvious downside to this approach is that whenever I upgrade airflow I
> have to double check in staging that my hack still works. But the function
> I patch is super simple and my change is like 4 extra lines of code, so I
> don't mind it.
>
> Chris
>
> On Sat, Aug 21, 2021, at 3:03 PM, Siddharth VP wrote:
>
> Ok. Before I give up on the idea, how so? To me it looks like a new
> feature entirely, so not sure what you mean by "changing fileloc like that".
>
> On Sat, 21 Aug 2021 at 22:28, Ash Berlin-Taylor <[email protected]> wrote:
>
> FYI: Changing fileloc like that will have un-intended consequences to
> execution in future versions (i.e. 2.2) so we can't do that.
>
> -ash
>
> On Sat, Aug 21 2021 at 09:50:57 +0530, Siddharth VP <[email protected]>
> wrote:
>
> Yes per my commit linked above, it would be the yaml file that is shown in
> the webserver code view (because the fileloc field points to that).
>
> I've been doing something similar to what Damian said, but with the
> difference that I've to generate the YAMLs programmatically based on
> parameters received at an API endpoint - and implement all the CRUD
> operations on these yaml config files. During POC testing, I saw that when
> there were 200+ unpaused DAGs generated this way, the dag processor timeout
> was being hit. On increasing that timeout to a higher value, I got a
> message on the UI that the last scheduler heartbeat was 1 minute ago (which
> is likely because the scheduler was busy with DAG processing for the whole
> minute).
>
> That's what brings me to this proposal. By embedding the task of parsing
> the yaml/json within Airflow, dynamic dags are supported in a much more
> "native" way, such that all timeouts and intervals apply individually to
> the config files.
>
> This won't replace dag-factory <https://github.com/ajbosco/dag-factory> and
> other ecosystem tools (because we still need to have the "builder" python
> code), rather improve the scalability when using such tools, and avoid
> compromising on scheduler availability (though I may have had this issue
> only because I was using 1.10.12).
>
> Would love to hear feedback on whether the patch is PR-worthy -- because I
> think it is quite simple (doesn't require any schema changes for instance)
> but still addresses a lot of dynamic workflow needs.
>
> On Sat, 21 Aug 2021 at 03:29, Jarek Potiuk <[email protected]> wrote:
>
> Agree with Ash here. It's OK to present different view of the "source" of
> the DAG once we parsed the Python code. This can be done and it could be as
> easy as
>
> a) adding a field to dag to point to a "definition file" if the DAGs are
> produced by parsing files from source folder
> b) API call/parameter to submit/fetch the dag (providing we implement some
> form of DAG fetcher/DAG submission)
>
> On Fri, Aug 20, 2021 at 10:49 PM Ash Berlin-Taylor <[email protected]> wrote:
>
> Changing the code view to show the YAML is now "relatively" easy to
> achieve, at least from the webserver point of view, as since 2.0 it doesn't
> read the files on disk, but from the DB.
>
> There's a lot of details, but changing the way these DagCode rows are
> written could be achievable whilst still keeping the "there must be a
> python file to generate the dag".
>
> -ash
>
> On Fri, Aug 20 2021 at 20:41:55 +0000, "Shaw, Damian P." <
> [email protected]> wrote:
>
> I’d personally find this very useful. There’s usually extra information I
> have about the DAG, and the current “docs_md” is usually not nearly
> sufficient enough as it’s poorly placed so if I start adding a lot of info
> it gets in the way of the regular UI. Also last I tested the markdown
> formatting didn’t work and neither did the other formatter options.
>
>
>
> But I’m not sure how much other people have demand for this.
>
>
>
> Thanks,
>
> Damian
>
>
>
> *From:* Collin McNulty <[email protected]>
> *Sent:* Friday, August 20, 2021 16:36
> *To:* [email protected]
> *Subject:* Re: [DISCUSS] Adding better support for parametrized DAGs and
> dynamic DAGs using JSON/YAML dataformats
>
>
>
> On the topic of pointing the code view to yaml, would we alternatively
> consider adding a view on the UI that would allow arbitrary text content?
> This could be accomplished by adding an optional parameter to the dag
> object that allowed you to pass text (or a filepath) that would then go
> through a renderer (e.g. markdown). It could be a readme, or yaml content
> or anything the author wanted.
>
>
>
> Collin
>
>
>
> On Fri, Aug 20, 2021 at 3:27 PM Shaw, Damian P. <
> [email protected]> wrote:
>
> FYI this is what I did on one of my past projects for Airflow.
>
>
>
> The users wanted to write their DAGs as YAML files so my “DAG file” was a
> Python script that read the YAML files and converted them to DAGs. It was
> very easy to do and worked because of the flexibility of Airflow.
>
>
>
> The one thing that would have been nice though is if I could of easily
> changed the “code view” in Airflow to point to the relevant YAML file
> instead of the less useful “DAG file”.
>
>
>
> Damian
>
>
>
> *From:* Jarek Potiuk <[email protected]>
> *Sent:* Friday, August 20, 2021 16:21
> *To:* [email protected]
> *Cc:* [email protected]
> *Subject:* Re: [DISCUSS] Adding better support for parametrized DAGs and
> dynamic DAGs using JSON/YAML dataformats
>
>
>
> Airflow DAGS are Python code.This is a very basic assumption - which is
> not likely to change. Ever.
>
>
>
> And we are working on making it even more powerful. Writing DAGs in
> yaml/json makes them less powerful and less flexible. This is fine if you
> want to build on top of airflow and build a more declarative way of
> defining dags and use airflow to run it under the hood.
>
> if you think there is a group of users who can benefit from that - cool.
> You can publish a code to convert those to Airflow DAGs and submit it to
> our Ecosystem page. There are plenty of tlike "CWL - Common Workflow
> Language" and others:
>
> https://airflow.apache.org/ecosystem/#tools-integrating-with-airflow
>
>
>
> J.
>
>
>
> On Fri, Aug 20, 2021 at 2:48 PM Siddharth VP <[email protected]>
> wrote:
>
> Have we considered allowing dags in json/yaml formats before? I came up
> with a rather straightforward way to address parametrized and dynamic DAGs
> in Airflow, which I think makes dynamic dags work at scale.
>
>
>
> *Background / Current limitations:*
>
> 1. Dynamic DAG generation using single-file methods
> <https://www.astronomer.io/guides/dynamically-generating-dags#single-file-methods>
>  can
> cause scalability issues
> <https://www.astronomer.io/guides/dynamically-generating-dags#scalability>
> where there are too many active DAGs per file. The
> dag_file_processor_timeout is applied to the loader file, so *all* dynamically
> generated dags need to be processed in that time. Sure the timeout could be
> increased, but that may be undesirable (what if there are other static DAGs
> in the system on which we really want to enforce a small timeout?)
>
> 2. Parametrizing DAGs in Airflow is difficult. There is no good way to
> have multiple workflows that differ only by choices of some constants.
> Using TriggerDagRunOperator to trigger a generic DAG with conf doesn't give
> a native-ish experience as it creates DagRuns of the *triggered* dag
> rather than *this* dag - which also means a single scheduler log file.
>
>
>
> *Suggested approach:*
>
> 1. User writes configuration files in JSON/YAML format. The schema can be
> arbitrary except for one condition that it must have a *builder* parameter
> with the path to a python file.
>
> 2. User writes the "builder" - a python file containing a make_dag method
> that receives the parsed json/yaml and returns a DAG object. (Just a
> sample strategy, we could instead say the file should contain a class that
> extends an abstract DagBuilder class.)
>
> 2. Airflow reads JSON/YAML files as well from the dags directory. It
> parses the file, imports the builder python file, and passes the parsed
> json/yaml to it and collects the generated DAG into the DagBag.
>
>
>
> *Sample implementation:*
>
> See
> https://github.com/siddharthvp/airflow/commit/47bad51fc4999737e9a300b134c04bbdbd04c88a;
> only major code change is in dagbag.py
>
>
>
> *Result:*
>
> Dag file processor logs show yaml/json file (instead of the builder python
> file). Each dynamically generated dag gets its own scheduler log file.
>
> The configs dag_dir_list_interval, min_file_process_interval,
> file_parsing_sort_mode all directly apply to dag config files.
>
> If the json/yaml fail to parse, it's registered as an import error.
>
>
>
> Would like to know your thoughts on this. Thanks!
>
> Siddharth VP
>
>
>
>
> --
>
> +48 660 796 129
>
>
>
>
> ==============================================================================
> Please access the attached hyperlink for an important electronic
> communications disclaimer:
> http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
>
> ==============================================================================
>
>
>
> ==============================================================================
> Please access the attached hyperlink for an important electronic
> communications disclaimer:
> http://www.credit-suisse.com/legal/en/disclaimer_email_ib.html
>
> ==============================================================================
>
>
>
> --
> +48 660 796 129
>
>
>

Reply via email to