chore(source-google-search-console): Migrate to CDK v6

[Christo Grabowski (ChristoGrab)]

What

Migrates google-search-console to the latest major CDK version, in order to facilitate the migration of streams from the legacy CDK to low-code.

User Impact

Hopefully none, if any should be positive in the form of more mature CDK logic.

Can this PR be safely reverted and rolled back?

• YES 💚
• NO ❌

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated (UTC)
airbyte-docs	✅ Ready (Inspect)	Visit Preview	💬 Add feedback	May 6, 2025 0:14am

[Christo Grabowski (ChristoGrab)]

The revert_migration test function was reformatting the local test_config.json file on each run, leading to formatting issues with our current pre-commit hooks. Not a big deal, but could be annoying for local dev work. Switching to a backup/restore of the original file fixed the issue for me.

[Christo Grabowski (ChristoGrab)]

Reverted because pytest 8.0 does not play nice with the pytest-lazy-fixture dependency which is used in unit tests. Guessing there's a decent chance that unit tests will have to be adjusted anyway as we're migrating streams, so maybe we can rip out this dependency soon, but feels out of scope for this PR.

[Christo Grabowski (ChristoGrab)]

Kicked off a regression test here. Might take a while to finish, this source has some beefy read times:

https://github.com/airbytehq/airbyte/actions/runs/14742276635/job/41382826417

[Maxime Carbonneau-Leclerc (maxi297)]

I saw from the regression tests that some of the streams have switched to being resumable. What are those streams? Should we disable this to keep the same behavior?

[Christo Grabowski (ChristoGrab)]

/format-fix

Format-fix job started... Check job output.

✅ Changes applied successfully. (d88bdbf)

[Christo Grabowski (ChristoGrab)]

I saw from the regression tests that some of the streams have switched to being resumable. What are those streams? Should we disable this to keep the same behavior?

Maxime Carbonneau-Leclerc (@maxi297) Was having a hard time interpreting the test report to figure out which streams were affected, but checking locally I noticed that sites and sitemaps streams had switched from false to true by default with the CDK bump. I've disabled the flag for those streams and will see if the diff still pops up in regression tests.

Test run url: https://github.com/airbytehq/airbyte/actions/runs/14762282210

[Maxime Carbonneau-Leclerc (maxi297)]

There seems to be some mismatches when syncing using a state. Are we able to explain these?

[Christo Grabowski (ChristoGrab)]

Maxime Carbonneau-Leclerc (@maxi297) I don't think the record diffs in search_analytics_keyword_page_report are too worrying as the reads are so long and the data could be getting updated. Definitely interesting that search_analytics_keyword_site_report_by_site didn't return any records though, running another round of regression tests just for that stream to make sure it wasn't a fluke.

Link here: https://github.com/airbytehq/airbyte/actions/runs/14781221599/job/41500384114

NOTE: tested locally and didn't see anything suspicious from search_analytics_keyword_site_report_by_site, it hit a 400 error that was handled gracefully and synced records as expected

[Christo Grabowski (ChristoGrab)]

There is definitely something up with the search_analytics_keyword_site_report_by_site stream, it's still returning no records in the target version... Everything seems fine when I run locally, I'll keep poking though. What's especially odd is there were no diffs in record count before I flipped the is_resumable flag for sites and sitemaps. I'm not clear on how that could have affected this stream, but will try to verify if it's the cause somehow.

[Christo Grabowski (ChristoGrab)]

Maxime Carbonneau-Leclerc (@maxi297) I looked into it and I think this may actually be a case of incremental behavior that was not previously working for search_analytics_keyword_site_report_by_site. Running an incremental read with state set to { "date": "2050-08-28" } returns no records as expected on the updated version, but on master 100 records are still read.

I say this with the confidence of someone who is wrong about these things 95% of the time, but I think it may be worth moving forward with a progressive rollout.

[Maxime Carbonneau-Leclerc (maxi297)]

If you think it's good to move forward with this, I'll approve!

[Christo Grabowski (ChristoGrab)]

/approve-regression-tests

Record mismatch in search_analytics_keyword_site_report_by_site seems to be related to incremental reads correctly skipping records with cursor values < stream state

Check job output.

✅ Approving regression tests