airflow-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jarek Potiuk <Jarek.Pot...@polidea.com>
Subject Re: [PROPOSAL][AIP-36 DAG Versioning]
Date Sun, 02 Aug 2020 07:47:39 GMT
Few points from my sid (and proposal!):

1) Agree with Max -  with a rather strong NO for pickles (however,
indeed cloudpickle solves some of the problems). Pickles came up in
our discussion in Polidea recently and the overall message was "no". I
agree with Max here - if we can ship python code, turning that into
pickle for transit makes little sense to me and brings a plethora of
problems.

2) I think indeed the versioning solution should treat the "DagRun"
structure atomically. While I see why we would like to go with the
UI/Scheduler only first rather than implementing them in the workers,
adding the "mixed version" is where it breaks down IMHO. Reasoning
about such "mixed version" dag is next to impossible. The current
behavior is not well defined and non-deterministic (depends on
scheduler delays, syncing, type of deployment, restarts of the works
etc.) we are moving it up to UI (thus users) rather than solving the
problem. So I am not a big fan of this and would rather solve it
"well" with atomicity.

3) I see the point of Dan as well - we had many discussions and many
times the idea about "submitting" the DAG for execution via the API
came up - and it makes sense IMHO.

Proposal: Implement full versioning with code shipping via DB wheels
BLOB (akin to serialized DAGs).

I understand that the big issue is how to actually "ship" the code to
the worker. And - maybe a wild idea - we can kill several birds with
the same stone.

There were plenty of discussions on how we could do that but one was
never truly explored - using wheel packages.

For those who do not know them, there is the PEP:
https://www.python.org/dev/peps/pep-0427/

Wheels allow to "package" python code in a standard way. They are
portable ("purelib" + contain .py rather than .pyc code), they have
metadata, versioning information, they can be signed for security,
They can contain other packages or python code, Why don't we let
scheduler to pack the fingerprinted version of the DAG in a .whl and
store it as a blob in a DB next to the serialized form?

There were concerns about the size of the code to keep in the DB - but
we already use the DB for serialized DAGs and it works fine (I believe
we only need to add compressing of the JSon serialized form - as we've
learned from AirBnb during their talk at the Airflow Summit - wheels
are already compressed). Also - each task will only need the
particular "version" of one DAG so even if we keep many of them in the
DB, the old version will pretty soon go "cold" and will never be
retrieved (and most DBs will handle it well with caching/indexes).

And if we want to add "callables" from other files - there is nothing
to stop the person who defines dag to add list of files that should be
packaged together with the main DAG file (additional_python_files =
["common/my_fantastic_library.py"] in DAG constructor). Or we could
auto-add all files after the DAG gets imported (i.e. package
automatically all files that are imported for that particular DAG from
the "dags" folder"). That should be rather easy.

This way we could ship the code to workers for the exact version that
the DagRun uses. And they can be cached and unpacked/installed to a
virtualenv for the execution of that single task. That should be super
quick. Such virtualenv can be wiped out after execution.

Then we got what Max wants (atomicity of DagRuns) and what Dan wants
(the API could simply accept "Wheel file + the Dag id". We have the
isolation between tasks running on the same worker (based on
virtualenv) so that each process in the same worker can run a
different version of the same Dag. We have much less confusion for the
UI.

Extra bonus 1: we can expand it to package different dependencies in
the wheels as well - so that if an operator requires a different
(newer) version of a python library, it could be packaged together
with the DAG in the same .whl file. This is also a highly requested
feature.
Extra bonus 2: workers will stop depending on the DAG file mount (!)
which was our long term goal and indeed as Dan mentioned - a great
step towards multi-tenancy.

J.






On Fri, Jul 31, 2020 at 6:41 AM Maxime Beauchemin
<maximebeauchemin@gmail.com> wrote:
>
> Having tried it early on, I'd advocate pretty strongly against pickles and
> would rather not get too deep into the why here. Short story is they can
> pull the entire memory space or much more than you want, and it's
> impossible to reason about where they end. For that reason and other
> reasons, they're a security issue. Oh and some objects are not picklable
> (Jinja templates! to name a problematic one...). I've also seen
> secret-related classes that raise when pickled (thank god!).
>
> About callback and other things like that, it's quite a puzzle in python.
> One solution would be to point to a python namespace
> callback="preset.airflow_utils.slack_callback" and assume the function has
> to exist in the remote interpreter. Personally I like the DagFetcher idea
> (it could be great to get a pointer to that mailing list thread here),
> specifically the GitDagFetcher. I don't know how [un]reasonable it is, but
> I hate pickles so much that shipping source code around seems much more
> reasonable to me. I think out there there's a talk from Mike Star about
> Dataswarm at FB and he may mention how their workers may git shallow clone
> the pipeline repo. Or maybe they use that "beautifully ugly" hack to use
> a gitfs fuse [file system in user space] on the worker [could get deeper
> into that, not sure how reasonable that is either].
>
> About fingerprints, a simple `start_date = datetime.now() - timedelta(1)`
> may lead to a never-repeating fingerprint. From memory the spec doesn't
> list out the properties considered to build the hash. It be helpful to
> specify and review that list.
>
> Max
>
> On Wed, Jul 29, 2020 at 5:20 AM Kaxil Naik <kaxilnaik@gmail.com> wrote:
>
> > Thanks, both Max and Dan for your comments, please check my reply below:
> >
> >
> > >  Personally I vote for a DAG version to be pinned and consistent for the
> > > duration of the DAG run. Some of the reasons why:
> > > - it's easier to reason about, and therefore visualize and troubleshoot
> > > - it prevents some cases where dependencies are never met
> > > - it prevents the explosion of artifact/metadata (one serialization per
> > > dagrun as opposed to one per scheduler cycle) in the case of a dynamic
> > DAG
> > > whose fingerprint is never the same.
> >
> >
> > In this AIP, we were only looking to fix the current "Viewing behaviour"
> > and
> > we were intentionally not changing the execution behaviour.
> > The change you are suggesting means we need to introduce DAG Versioning for
> > the
> > workers too. This will need more work as can't use the Serialised
> > Representation
> > to run the task since users could use custom modules in a different part of
> > code,
> > example the PythonOperator has python_callable that allows running any
> > arbitrary code.
> > A similar case is with the *on_*_callbacks* defined on DAG.
> >
> > Based on the current scope of the AIP, we still plan to use the actual DAG
> > files for the
> > execution and not use Serialized DAGs for the workers.
> >
> > To account for all the custom modules we will have to start looking at
> > pickle (cloudpickle).
> >
> > I'm certain that there are lots of
> > > those DAGs out there, and that it will overwhelm the metadata database,
> > and
> > > confuse the users. For an hourly DAG is would mean 24 artifact per day
> > > instead of 1000+
> >
> >
> > What kind of dynamic DAGs are we talking about here, I would think the DAG
> > signature won't change
> > but I might be wrong, can you give an example, please.
> >
> > If backwards compatibility in behavior is a concern, I'd recommend adding a
> > > flag to the DAG class and/or config and make sure we're doing the right
> > > thing by default. People who want backward compatibility would have to
> > > change that default. But again, that's a lot of extra and confusing
> > > complexity that will likely be the source of bugs and user confusion.
> > > Having a clear, easy to reason about execution model is super important.
> >
> > Think about visualizing a DAG that shapeshifted 5 times during its
> > > execution, how does anyone make sense of that?
> >
> >
> > Wouldn't that be an edge case? How often would someone change the DAG
> > structure in the middle of
> > a DAG execution. And since if they do change, the Graph View should show
> > all the tasks that were
> > run, if it just shows based on the latest version, the behaviour would be
> > the same as now.
> >
> > --------
> >
> > Strongly agree with Max's points, also I feel the right way to go about
> > > this is instead of Airflow schedulers/webservers/workers reading DAG
> > Python
> > > files, they would instead read from serialized representations of the
> > DAGs
> > > (e.g. json representation in the Airflow DB). Instead of DAG owners
> > pushing
> > > their DAG files to the Airflow components via varying mechanisms (e.g.
> > > git), they would instead call an Airflow CLI to push the serialized DAG
> > > representations to the DB, and for things like dynamic DAGs you could
> > > populate them from a DAG or another service.
> >
> >
> > Airflow Webserver and the Scheduler will definitely read from the
> > Serialized representation as
> > they don't need all the code from the DAG files.
> >
> > While the workers definitely need access to DAG files as the
> > tasks/operators would be using
> > code form custom modules and classes which are required to run the tasks.
> >
> > If we do want to go down that route we will have to use something like
> > cloudpickle that serializes
> > entire DAG file and their dependencies. And also ensure that someone is not
> > able to change the pickled
> > source when sending from executor to the worker as that poses a big
> > security risk.
> >
> > - Kaxil
> >
> > On Wed, Jul 29, 2020 at 12:43 PM Jacob Ward <jward@brandwatch.com> wrote:
> >
> > > I came here to say what Max has said, only less eloquently.
> > >
> > > I do have one concern with locking the version for a single run.
> > Currently
> > > it is possible for a user to create a dag which intentionally changes as
> > a
> > > dag executes, i.e. dynamically creating a task for the dag during a run
> > by
> > > modifying external data, but this change would prevent that. I'm of the
> > > opinion that this situation is bad practice anyway so it doesn't matter
> > if
> > > we make it impossible to do, but others may disagree.
> > >
> > > On Tue, 28 Jul 2020 at 17:08, Dan Davydov <ddavydov@twitter.com.invalid>
> > > wrote:
> > >
> > > > Strongly agree with Max's points, also I feel the right way to go about
> > > > this is instead of Airflow schedulers/webservers/workers reading DAG
> > > Python
> > > > files, they would instead read from serialized representations of the
> > > DAGs
> > > > (e.g. json representation in the Airflow DB). Instead of DAG owners
> > > pushing
> > > > their DAG files to the Airflow components via varying mechanisms (e.g.
> > > > git), they would instead call an Airflow CLI to push the serialized DAG
> > > > representations to the DB, and for things like dynamic DAGs you could
> > > > populate them from a DAG or another service.
> > > >
> > > > This would also enable other features like stronger
> > > security/multi-tenancy.
> > > >
> > > > On Tue, Jul 28, 2020 at 6:44 PM Maxime Beauchemin <
> > > > maximebeauchemin@gmail.com> wrote:
> > > >
> > > > > > "mixed version"
> > > > >
> > > > > Personally I vote for a DAG version to be pinned and consistent for
> > the
> > > > > duration of the DAG run. Some of the reasons why:
> > > > > - it's easier to reason about, and therefore visualize and
> > troubleshoot
> > > > > - it prevents some cases where dependencies are never met
> > > > > - it prevents the explosion of artifact/metadata (one serialization
> > per
> > > > > dagrun as opposed to one per scheduler cycle) in the case of a
> > dynamic
> > > > DAG
> > > > > whose fingerprint is never the same. I'm certain that there are lots
> > of
> > > > > those DAGs out there, and that it will overwhelm the metadata
> > database,
> > > > and
> > > > > confuse the users. For an hourly DAG is would mean 24 artifact per
> > day
> > > > > instead of 1000+
> > > > >
> > > > > If backwards compatibility in behavior is a concern, I'd recommend
> > > > adding a
> > > > > flag to the DAG class and/or config and make sure we're doing the
> > right
> > > > > thing by default. People who want backward compatibility would have
> > to
> > > > > change that default. But again, that's a lot of extra and confusing
> > > > > complexity that will likely be the source of bugs and user confusion.
> > > > > Having a clear, easy to reason about execution model is super
> > > important.
> > > > >
> > > > > Think about visualizing a DAG that shapeshifted 5 times during its
> > > > > execution, how does anyone make sense of that?
> > > > >
> > > > > Max
> > > > >
> > > > > On Tue, Jul 28, 2020 at 3:14 AM Kaxil Naik <kaxilnaik@gmail.com>
> > > wrote:
> > > > >
> > > > > > Thanks Max for your comments.
> > > > > >
> > > > > >
> > > > > > *DAG Fingerprinting: *this can be tricky, especially in regards
to
> > > > > dynamic
> > > > > > > DAGs, where in some cases each parsing of the DAG can result
in a
> > > > > > different
> > > > > > > fingerprint. I think DAG and tasks attributes are left
out from
> > the
> > > > > > > proposal that should be considered as part of the fingerprint,
> > like
> > > > > > trigger
> > > > > > > rules or task start/end datetime. We should do a full pass
of all
> > > DAG
> > > > > > > arguments and make sure we're not forgetting anything that
can
> > > change
> > > > > > > scheduling logic. Also, let's be careful that something
as simple
> > > as
> > > > a
> > > > > > > dynamic start or end date on a task could lead to a different
> > > version
> > > > > > each
> > > > > > > time you parse.
> > > > > >
> > > > > >
> > > > > >
> > > > > > The short version of Dag Fingerprinting would be
> > > > > > just a hash of the Serialized DAG.
> > > > > >
> > > > > > *Example DAG*: https://imgur.com/TVuoN3p
> > > > > > *Example Serialized DAG*: https://imgur.com/LmA2Bpr
> > > > > >
> > > > > > It contains all the task & DAG parameters. When they change,
> > > Scheduler
> > > > > > writes
> > > > > > a new version of Serialized DAGs to the DB. The Webserver then
> > reads
> > > > the
> > > > > > DAGs from the DB.
> > > > > >
> > > > > > I'd recommend limiting serialization/storage of one version
> > > > > > > per DAG Run, as opposed to potentially everytime the DAG
is
> > parsed
> > > -
> > > > > once
> > > > > > > the version for a DAG run is pinned, fingerprinting is
not
> > > > re-evaluated
> > > > > > > until the next DAG run is ready to get created.
> > > > > >
> > > > > >
> > > > > > This is to handle Scenario 3 where a DAG structure is changed
> > > mid-way.
> > > > > > Since we don't intend to
> > > > > > change the execution behaviour, if we limit Storage of 1 version
> > per
> > > > DAG,
> > > > > > it won't actually show what
> > > > > > was run.
> > > > > >
> > > > > > Example Dag v1: Task A -> Task B -> Task C
> > > > > > The worker has completed the execution of Task B and is just
about
> > to
> > > > > > complete the execution of Task B.
> > > > > >
> > > > > > The 2nd version of DAG is deployed: Task A -> Task D
> > > > > > Now Scheduler queued Task D and it will run to completion. (Task
C
> > > > won't
> > > > > > run)
> > > > > >
> > > > > > In this case, "the actual representation of the DAG" that run
is
> > > > neither
> > > > > v1
> > > > > > nor v2 but a "mixed version"
> > > > > >  (Task A -> Task B -> Task D). The plan is that the Scheduler
will
> > > > create
> > > > > > this "mixed version" based on what ran
> > > > > > and the Graph View would show this "mixed version".
> > > > > >
> > > > > > There would also be a toggle button on the Graph View to select
v1
> > or
> > > > v2
> > > > > > where the tasks will be highlighted to show
> > > > > > that a particular task was in v1 or v2 as shown in
> > > > > >
> > > > > >
> > > > >
> > > >
> > >
> > https://cwiki.apache.org/confluence/download/attachments/158868919/Picture%201.png?version=2&modificationDate=1595612863000&api=v2
> > > > > >
> > > > > >
> > > > > >
> > > > > > *Visualizing change in the tree view:* I think this is very
complex
> > > and
> > > > > > > many things can make this view impossible to render (task
> > > dependency
> > > > > > > reversal, cycles across versions, ...). Maybe a better
visual
> > > > approach
> > > > > > > would be to render independent, individual tree views for
each
> > DAG
> > > > > > version
> > > > > > > (side by side), and doing best effort aligning the tasks
across
> > > > blocks
> > > > > > and
> > > > > > > "linking" tasks with lines across blocks when necessary.
> > > > > >
> > > > > >
> > > > > > Agreed, the plan is to do the best effort aligning.
> > > > > > At this point in time, task additions to the end of the DAG
are
> > > > expected
> > > > > to
> > > > > > be compatible,
> > > > > > but changes to task structure within the DAG may cause the tree
> > view
> > > > not
> > > > > to
> > > > > > incorporate “old” and “new” in the same view, hence
that won't be
> > > > shown.
> > > > > >
> > > > > > Regards,
> > > > > > Kaxil
> > > > > >
> > > > > > On Mon, Jul 27, 2020 at 6:02 PM Maxime Beauchemin <
> > > > > > maximebeauchemin@gmail.com> wrote:
> > > > > >
> > > > > > > Some notes and ideas:
> > > > > > >
> > > > > > > *DAG Fingerprinting: *this can be tricky, especially in
regards
> > to
> > > > > > dynamic
> > > > > > > DAGs, where in some cases each parsing of the DAG can result
in a
> > > > > > different
> > > > > > > fingerprint. I think DAG and tasks attributes are left
out from
> > the
> > > > > > > proposal that should be considered as part of the fingerprint,
> > like
> > > > > > trigger
> > > > > > > rules or task start/end datetime. We should do a full pass
of all
> > > DAG
> > > > > > > arguments and make sure we're not forgetting anything that
can
> > > change
> > > > > > > scheduling logic. Also, let's be careful that something
as simple
> > > as
> > > > a
> > > > > > > dynamic start or end date on a task could lead to a different
> > > version
> > > > > > each
> > > > > > > time you parse. I'd recommend limiting serialization/storage
of
> > one
> > > > > > version
> > > > > > > per DAG Run, as opposed to potentially everytime the DAG
is
> > parsed
> > > -
> > > > > once
> > > > > > > the version for a DAG run is pinned, fingerprinting is
not
> > > > re-evaluated
> > > > > > > until the next DAG run is ready to get created.
> > > > > > >
> > > > > > > *Visualizing change in the tree view:* I think this is
very
> > complex
> > > > and
> > > > > > > many things can make this view impossible to render (task
> > > dependency
> > > > > > > reversal, cycles across versions, ...). Maybe a better
visual
> > > > approach
> > > > > > > would be to render independent, individual tree views for
each
> > DAG
> > > > > > version
> > > > > > > (side by side), and doing best effort aligning the tasks
across
> > > > blocks
> > > > > > and
> > > > > > > "linking" tasks with lines across blocks when necessary.
> > > > > > >
> > > > > > > On Fri, Jul 24, 2020 at 12:46 PM Vikram Koka <
> > vikram@astronomer.io
> > > >
> > > > > > wrote:
> > > > > > >
> > > > > > > > Team,
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > We just created 'AIP-36 DAG Versioning' on Confluence
and would
> > > > very
> > > > > > much
> > > > > > > > appreciate feedback and suggestions from the community.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> > https://cwiki.apache.org/confluence/display/AIRFLOW/AIP-36+DAG+Versioning
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > The DAG Versioning concept has been discussed on multiple
> > > occasions
> > > > > in
> > > > > > > the
> > > > > > > > past and has been a topic highlighted as part of Airflow
2.0 as
> > > > well.
> > > > > > We
> > > > > > > at
> > > > > > > > Astronomer have heard data engineers at several enterprises
ask
> > > > about
> > > > > > > this
> > > > > > > > feature as well, for easier debugging when changes
are made to
> > > DAGs
> > > > > as
> > > > > > a
> > > > > > > > result of evolving business needs.
> > > > > > > >
> > > > > > > >
> > > > > > > > As described in the AIP, we have a proposal focused
on ensuring
> > > > that
> > > > > > the
> > > > > > > > visibility behaviour of Airflow is correct, without
changing
> > the
> > > > > > > execution
> > > > > > > > behaviour. We considered changing the execution behaviour
as
> > > well,
> > > > > but
> > > > > > > > decided that the risks in changing execution behavior
were too
> > > high
> > > > > as
> > > > > > > > compared to the benefits and therefore decided to
limit the
> > scope
> > > > to
> > > > > > only
> > > > > > > > making sure that the visibility was correct.
> > > > > > > >
> > > > > > > >
> > > > > > > > We would like to attempt this based on our experience
running
> > > > Airflow
> > > > > > as
> > > > > > > a
> > > > > > > > service. We believe that this benefits Airflow as
a project and
> > > the
> > > > > > > > development experience of data engineers using Airflow
across
> > the
> > > > > > world.
> > > > > > > >
> > > > > > > >
> > > > > > > >  Any feedback, suggestions, and comments would be
greatly
> > > > > appreciated.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > Best Regards,
> > > > > > > >
> > > > > > > >
> > > > > > > > Kaxil Naik, Ryan Hamilton, Ash Berlin-Taylor, and
Vikram Koka
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> > >
> > > --
> > >
> > > Jacob Ward    |    Graduate Data Infrastructure Engineer
> > >
> > > jward@brandwatch.com
> > >
> > >
> > > NEW YORK   | BOSTON   | BRIGHTON   | LONDON   | BERLIN |   STUTTGART |
> > > PARIS   | SINGAPORE | SYDNEY
> > >
> >



--

Jarek Potiuk
Polidea | Principal Software Engineer

M: +48 660 796 129

Mime
View raw message