What this PR does / why we need it:

Problem

CouchbaseOnlineStore.online_write_batch wraps each upsert in except Exception, logs it, and carries on. progress(1) still runs for the row that failed. So when Couchbase times out or rejects a write, materialization finishes normally, reports every row as written, and the online store is missing those rows. update() does the same when a scope or collection can't be created, so feast apply also succeeds against a store that isn't set up.

The rest of the file already draws the line in the right place. online_read catches only DocumentNotFoundException, and update() already catches ScopeAlreadyExistsException and CollectionAlreadyExistsException on their own. Of the other 22 online stores, 21 have no handler around their writes and MySQL catches and re-raises, so a failed write reaches the caller everywhere else.

Reproduction

A mocked collection whose upsert raises TimeoutException, on master. online_write_batch with three rows logs three errors and returns normally:

ERROR: Error upserting document test_project:driver_stats:...:trips: TimeoutException(<message=upsert timed out>)
ERROR: Error upserting document test_project:driver_stats:...:trips: TimeoutException(<message=upsert timed out>)
ERROR: Error upserting document test_project:driver_stats:...:trips: TimeoutException(<message=upsert timed out>)

No exception reaches the caller, and progress is called three times.

Change

The three generic handlers keep their log line and raise. The "already exists" handlers and the read path are unchanged, so re-running feast apply against an existing scope and collection still works.

Verification

New tests/unit/infra/online_store/test_couchbase_online_store.py, 5 tests. With the source reverted to master and the tests kept, 3 fail: the write, the scope and the collection cases. The 2 that pass are the controls, a normal write and an already-existing scope and collection, which behave the same before and after.

The rest of tests/unit/infra/online_store gives the same result with and without this change. ruff check and ruff format --check are clean.

Which issue(s) this PR fixes:

None filed. The reproduction is above.

Checks

• I've made sure the tests are passing.
• My commits are signed off (git commit -s)
• My PR title follows conventional commits format

Testing Strategy

• Unit tests
• Integration tests
• Manual tests
• Testing is not required for this change

Misc

The store is marked alpha, so I kept this to the error handling and left everything else alone.