🎉 🎉 Source FB Marketing: performance and reliability fixes #9805

keu · 2022-01-26T05:12:03Z

What

this PR is based on #8385

It focused on several improvements.
Stability:

use campaign-level insights when account-level insights start to fail.
retry HTTP requests on a much lower level now, avoid unhandled fails during batch and pagination.
respect throttle and call rate limit
save cursor for each slice individually, thus making progress in sync even if one job fails, eventually, we will sync that job.

Performance:

check the status of running jobs with a single batch request.
deprecated lookback_window param because it actually used in the wrong way before, and really useless. FB updates insights after up to 28 days. Make it constant and read last 28 days (not from the current cursor, but from now() datetime)
track individual slices in the state, we process jobs ASAP and don't wait and are stuck in a single job, probably the biggest improvement.

Other fixes:

use correct logic for start_day, end_day, and cursor_value
fixed custom insights schema validation, now use pydantic
test all streams in SAT
raised code coverage from 60% to 75% (and more after merge of ( CDK: add support for streams with state attribute #9746 and SAT: Introduce new way to compare records based on PK #9718 #9768)

How

AsyncJob is responsible for syncing fixed time intervals now. I introduced a new class ParentAsyncJob to group smaller AsyncJobs. When AsyncJob fails for the first time it will try to restart, if it fails again InsightAsyncJobManager will split such job into group of smaller AsyncJobs, one job for each campaign active for a specific interval. We use 28day window to fetch IDs for these campaigns. The fail situation is quite rare, so this will degrade performance just a bit, but will improve reliability, it potentially could split further and use AdSets or even Ads to fetch data in smaller batches.
Especially it is important for date intervals bigger then 1 day (there are feature requests already for this)

🚨 User Impact 🚨

We deprecate day_per_job parameter, as it is not needed after the introduction of the split algorithm.
We deprecate lookback_window parameter because it was used in the wrong way before, and FB has a constant value for it.

Pre-merge Checklist

Expand the relevant checklist and delete the others.

New Connector

Community member or Airbyter

Airbyter

If this is a community PR, the Airbyte engineer reviewing this PR is responsible for the below items.

Create a non-forked branch based on this PR and test the below items on it
Build is successful
Credentials added to Github CI. Instructions.
/test connector=connectors/<name> command is passing.
New Connector version released on Dockerhub by running the /publish command described here
After the connector is published, connector added to connector index as described here
Seed specs have been re-generated by building the platform and committing the changes to the seed spec files, as described here

Updating a connector

Community member or Airbyter

Grant edit access to maintainers (instructions)
Secrets in the connector's spec are annotated with airbyte_secret
Unit & integration tests added and passing. Community members, please provide proof of success locally e.g: screenshot or copy-paste unit, integration, and acceptance test output. To run acceptance tests for a Python connector, follow instructions in the README. For java connectors run ./gradlew :airbyte-integrations:connectors:<name>:integrationTest.
Code reviews completed
Documentation updated
- Connector's README.md
- Connector's bootstrap.md. See description and examples
- Changelog updated in docs/integrations/<source or destination>/<name>.md including changelog. See changelog example
PR name follows PR naming conventions

Airbyter

If this is a community PR, the Airbyte engineer reviewing this PR is responsible for the below items.

Create a non-forked branch based on this PR and test the below items on it
Build is successful
Credentials added to Github CI. Instructions.
/test connector=connectors/<name> command is passing.
New Connector version released on Dockerhub by running the /publish command described here
After the new connector version is published, connector version bumped in the seed directory as described here
Seed specs have been re-generated by building the platform and committing the changes to the seed spec files, as described here

Connector Generator

Issue acceptance criteria met
PR name follows PR naming conventions
If adding a new generator, add it to the list of scaffold modules being tested
The generator test modules (all connectors with -scaffold in their name) have been updated with the latest scaffold by running ./gradlew :airbyte-integrations:connector-templates:generator:testScaffoldTemplates then checking in your changes
Documentation which references the generator is updated as needed.

Bump version

fix SAT

keu · 2022-02-12T01:59:15Z

/test connector=connectors/source-facebook-marketing

🕑 connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1832417116
❌ connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1832417116
🐛 https://gradle.com/s/iypgu5ude5lqa
Python short test summary info:

=========================== short test summary info ============================
FAILED test_core.py::TestSpec::test_match_expected[inputs0] - AssertionError:...
FAILED test_incremental.py::TestIncremental::test_two_sequential_reads[inputs0]
=================== 2 failed, 20 passed in 521.00s (0:08:40) ===================

keu · 2022-02-13T21:56:43Z

/test connector=connectors/source-facebook-marketing

🕑 connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1838093837
❌ connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1838093837
🐛 https://gradle.com/s/utg5ortl7m64w
Python short test summary info:

=========================== short test summary info ============================
FAILED test_incremental.py::TestIncremental::test_two_sequential_reads[inputs0]
=================== 1 failed, 21 passed in 434.52s (0:07:14) ===================

keu · 2022-02-13T22:25:25Z

/test connector=connectors/source-facebook-marketing

🕑 connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1838093837
❌ connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1838093837
🐛 https://gradle.com/s/utg5ortl7m64w
Python short test summary info:
=========================== short test summary info ============================
FAILED test_incremental.py::TestIncremental::test_two_sequential_reads[inputs0]
=================== 1 failed, 21 passed in 434.52s (0:07:14) ===================

incremental behaviour of Images is broken until #9746

sergei-solonitcyn · 2022-02-01T07:11:37Z

airbyte-integrations/connectors/source-facebook-marketing/source_facebook_marketing/source.py

        """Connection check to validate that the user-provided config can be used to connect to the underlying API

        :param config:  the user-input config object conforming to the connector's spec.json
-        :param logger:  logger object
+        :param _logger:  logger object


Seems like we don't use it anywhere. Shouldn't we also update line 46?

keu · 2022-02-16T20:31:19Z

/publish connector=source-facebook-marketing

❌ source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1855080691

keu · 2022-02-17T03:50:08Z

/publish connector=connectors/source-facebook-marketing

🕑 connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1856646105
✅ connectors/source-facebook-marketing https://github.com/airbytehq/airbyte/actions/runs/1856646105

…rmance-8282 # Conflicts: # airbyte-config/init/src/main/resources/config/STANDARD_SOURCE_DEFINITION/e7778cfc-e97c-4458-9ecb-b4f2bba8946c.json

vladimir-remar · 2022-04-26T14:20:27Z

@keu the possibility to set up an attribution window was really healpful for us. We understand the reason why you made it a constant, but our use case is another one. We are only using some data that will never change regardless of the attribution window (e.g., spend) and not using any of the data that could change (e.g., conversions). For this reason, incremental synch with a window of 3 days was more than enough for us. With this new approach, we are forced to retrieve a lot of data we don't need and with the constrains of the FB API, this connection takes 5 hours to run every day, while with 3 days attribution window would be only 20-30 mins. Would you be open to add a variable incremental_lookback_window with default 28 days? It would help our use case. I can do the PR.

sherifnada · 2022-04-26T19:31:08Z

@vladimir-remar could you propose a draft PR for what the spec would look like?

vladimir-remar · 2022-04-27T07:05:03Z

@vladimir-remar could you propose a draft PR for what the spec would look like?

@sherifnada Thanks for the answer and I will do it ASAP.

Dmytro Rezchykov and others added 30 commits December 3, 2021 12:08

Facebook Marketing performance improvement

57b47cb

add comments and little refactoring

69b2958

Bump version

fix after performance test

440fc6e

Check job status in advance, some other changes.

dff2421

Bump rc version

4112279

fix usage of AsyncJob methods, correct batch job retry

939d646

bump dev version

0f8439d

revert extra diff with master

86ce7db

Merge remote-tracking branch 'origin/master' into drezchykov/fb-perf

fa650b8

fix bad merge

d8f57d6

isort

5c7e18f

fix flake

c9deade

fix unit tests

07acea5

fix integration tests with the new config

213ac36

improve job status handling, limit concurrency to 10

93a03e1

WIP

47ad837

fix campaign jobs, refactor manager

483f5cd

big refactoring of async jobs, support random order of slices

4faabcb

update source _read_incremental to hook new state logic

ea1ac91

fix issues with timeout

c7ca0e3

remove debugging and clean up, improve retry logic

f66ce53

Merge remote-tracking branch 'origin/master' into drezchykov/fb-perf

cc5baea

merge changes from #8234

218db5c

fix call super _read_increment

2e56c22

generalize batch execution, add use_batch flag

8d560d8

fix SAT

fixing unittests

3c0754d

fix integration tests

76b529c

fix tests for AsyncJob

ca42a4e

fix unit_tests and batch yield

15e73f1

fix SAT

d09994f

format

c4e8ce2

keu temporarily deployed to more-secrets February 12, 2022 01:50 Inactive

fix reversed incremental stream

2d04abd

keu temporarily deployed to more-secrets February 12, 2022 02:00 Inactive

octavia-squidington-iii temporarily deployed to more-secrets February 12, 2022 02:01 Inactive

update spec.json for SAT

5101e85

keu temporarily deployed to more-secrets February 13, 2022 21:56 Inactive

octavia-squidington-iii temporarily deployed to more-secrets February 13, 2022 21:59 Inactive

sherifnada approved these changes Feb 14, 2022

View reviewed changes

sergei-solonitcyn approved these changes Feb 16, 2022

View reviewed changes

upgrade CDK and bump version

3d92326

keu temporarily deployed to more-secrets February 16, 2022 20:24 Inactive

octavia-squidington-iii temporarily deployed to more-secrets February 16, 2022 20:33 Inactive

octavia-squidington-iii temporarily deployed to more-secrets February 17, 2022 03:52 Inactive

eugene-kulak added 2 commits February 17, 2022 08:31

bump version

2e119a8

Merge remote-tracking branch 'origin/master' into keu/source-fb-perfo…

6965ef4

…rmance-8282 # Conflicts: # airbyte-config/init/src/main/resources/config/STANDARD_SOURCE_DEFINITION/e7778cfc-e97c-4458-9ecb-b4f2bba8946c.json

keu temporarily deployed to more-secrets February 17, 2022 04:35 Inactive

keu merged commit a3aae80 into master Feb 17, 2022

keu temporarily deployed to more-secrets February 17, 2022 04:35 Inactive

keu deleted the keu/source-fb-performance-8282 branch February 17, 2022 04:35

octavia-squidington-iii mentioned this pull request Feb 17, 2022

Bump Airbyte version from 0.35.30-alpha to 0.35.31-alpha #10446

Merged

vladimir-remar mentioned this pull request Apr 27, 2022

Source Facebook Marketing: Add lookback window to insights streams #12402

Merged

14 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

🎉 🎉 Source FB Marketing: performance and reliability fixes #9805

🎉 🎉 Source FB Marketing: performance and reliability fixes #9805

keu commented Jan 26, 2022 •

edited

Loading

keu commented Feb 12, 2022 •

edited by github-actions bot

Loading

keu commented Feb 13, 2022 •

edited by github-actions bot

Loading

keu commented Feb 13, 2022

sergei-solonitcyn Feb 1, 2022

keu commented Feb 16, 2022 •

edited by github-actions bot

Loading

keu commented Feb 17, 2022 •

edited by github-actions bot

Loading

vladimir-remar commented Apr 26, 2022

sherifnada commented Apr 26, 2022

vladimir-remar commented Apr 27, 2022

🎉 🎉 Source FB Marketing: performance and reliability fixes #9805

🎉 🎉 Source FB Marketing: performance and reliability fixes #9805

Conversation

keu commented Jan 26, 2022 • edited Loading

What

How

Recommended reading order

🚨 User Impact 🚨

Pre-merge Checklist

Community member or Airbyter

Airbyter

Community member or Airbyter

Airbyter

keu commented Feb 12, 2022 • edited by github-actions bot Loading

keu commented Feb 13, 2022 • edited by github-actions bot Loading

keu commented Feb 13, 2022

sergei-solonitcyn Feb 1, 2022

Choose a reason for hiding this comment

keu commented Feb 16, 2022 • edited by github-actions bot Loading

keu commented Feb 17, 2022 • edited by github-actions bot Loading

vladimir-remar commented Apr 26, 2022

sherifnada commented Apr 26, 2022

vladimir-remar commented Apr 27, 2022

keu commented Jan 26, 2022 •

edited

Loading

keu commented Feb 12, 2022 •

edited by github-actions bot

Loading

keu commented Feb 13, 2022 •

edited by github-actions bot

Loading

keu commented Feb 16, 2022 •

edited by github-actions bot

Loading

keu commented Feb 17, 2022 •

edited by github-actions bot

Loading