damccorm opened a new pull request, #41231: URL: https://github.com/apache/beam/pull/41231
### Summary In `google-cloud-bigtable >= 2.44.0` (upgraded in #40042 / #40960), `DirectRow.commit()` delegates to `_DataApiTarget.mutate_row(row_key, mutations)`, which only retries transient `DeadlineExceeded` and `ServiceUnavailable` errors when `all(m.is_idempotent() for m in mutations)` is `True`. When `DirectRow.set_cell()` is called without an explicit `timestamp`, it sets `timestamp_micros = -1` (`_SERVER_SIDE_TIMESTAMP`), causing `SetCell.is_idempotent()` to return `False` and disabling retries completely. If the single 20-second `MutateRow` attempt times out while a newly created Bigtable instance/table/column family is warming up (as seen in `TestWriteToBigtableXlangIT.test_delete_cells_mutation` in [run 43118590693](https://github.com/apache/beam/actions/runs/43118590693/job/130852919019)), `DirectRow.commit()` catches `GoogleAPICallError`, clears `self._mutations` in `finally`, and returns a non-OK `Status` without raising an exception—leaving the table empty and causing `actual_row = self.table.read_row('key-1')` to return `None` (`AttributeError: 'NoneType' object has no attribute 'find_cells'`). ### Changes - Pass an explicit `datetime.now(timezone.utc)` timestamp to `set_cell()` in pre-test setup writes in `bigtableio_it_test.py` so `SetCell.is_idempotent()` is `True` and `DirectRow.commit()` retries transient `DeadlineExceeded`/`ServiceUnavailable` errors during Bigtable instance/table warmup. - Assert `row.commit().code == 0` on pre-test setup commits so any setup failure surfaces immediately with its gRPC status code. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
