-
Notifications
You must be signed in to change notification settings - Fork 30
fix: handle backoff_strategies in CompositeErrorHandler #225
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
unit_tests/sources/declarative/requesters/error_handlers/test_composite_error_handler.py
Show resolved
Hide resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can we link the GitHub issue this change is related to?
📝 WalkthroughWalkthroughThe pull request introduces an enhancement to the Changes
Assessment against linked issues
The changes directly address the issue of configuring error handlers for backoff strategies, providing more flexibility in handling retryable errors. The implementation allows for custom backoff strategies to be properly configured and used, which was not working in previous versions. Wdyt about the clarity of the new tests? They seem to cover a range of scenarios effectively! 😊 ✨ Finishing Touches
Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actionable comments posted: 0
🧹 Nitpick comments (2)
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py (1)
82-96
: LGTM! The backoff_strategies implementation looks solid.The implementation correctly aggregates strategies from all handlers and handles edge cases well. The docstring is particularly helpful in explaining the current behavior and limitations.
Just a thought - would it make sense to add a debug log when no strategies are found, to help with troubleshooting? wdyt?
unit_tests/sources/declarative/requesters/test_http_requester.py (1)
908-936
: Integration test looks good! Addresses the previous review feedback.The test effectively validates that backoff strategies from the manifest are respected during HTTP requests. Nice job following up on the previous review comment about manifest integration!
One suggestion - would it be helpful to also verify the actual backoff time used? For example, checking that the 0.1s constant backoff was applied? wdyt?
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (3)
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py
(2 hunks)unit_tests/sources/declarative/requesters/error_handlers/test_composite_error_handler.py
(2 hunks)unit_tests/sources/declarative/requesters/test_http_requester.py
(3 hunks)
⏰ Context from checks skipped due to timeout of 90000ms (7)
- GitHub Check: Check: 'source-pokeapi' (skip=false)
- GitHub Check: Check: 'source-the-guardian-api' (skip=false)
- GitHub Check: Check: 'source-shopify' (skip=false)
- GitHub Check: Check: 'source-hardcoded-records' (skip=false)
- GitHub Check: Pytest (All, Python 3.11, Ubuntu)
- GitHub Check: Pytest (All, Python 3.10, Ubuntu)
- GitHub Check: Pytest (Fast)
🔇 Additional comments (2)
unit_tests/sources/declarative/requesters/error_handlers/test_composite_error_handler.py (2)
280-314
: Great test coverage! The parameterized tests look comprehensive.The test cases effectively cover all important scenarios including empty, single, and multiple handler configurations.
317-351
: The strategy prioritization test is well thought out.Good job testing that the first handler's strategy is respected even when the second handler's response filter is triggered. This is crucial for validating the behavior described in the docstring.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM! Can we create an issue to improve on backoff strategies and fix the issue when there are multiple error handlers in the composite error handler?
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py
Outdated
Show resolved
Hide resolved
Created the issue to improve the handling of backoff strategies, linking here for additional visibility: https://github.com/airbytehq/airbyte-internal-issues/issues/11449 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actionable comments posted: 0
🧹 Nitpick comments (2)
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py (2)
82-101
: Great implementation! Consider adding a warning for multiple strategies?The implementation looks solid and the docstring is very informative about the "first strategy wins" behavior. Since subsequent strategies are ignored, what do you think about adding a warning when multiple strategies are present? This could help users understand why their additional strategies aren't being used. wdyt?
Here's a potential enhancement:
@property def backoff_strategies(self) -> Optional[List[BackoffStrategy]]: all_strategies = [] for handler in self.error_handlers: if hasattr(handler, "backoff_strategies") and handler.backoff_strategies: all_strategies.extend(handler.backoff_strategies) + if len(all_strategies) > 1: + import warnings + warnings.warn( + "Multiple backoff strategies detected. Only the first strategy will be used, subsequent strategies will be ignored.", + UserWarning + ) return all_strategies if all_strategies else None
84-96
: Enhance return type documentation?The docstring is excellent in explaining the behavior! Would you consider adding a Returns section that explicitly documents the return type and its structure? Something like:
""" Combines backoff strategies from all child error handlers into a single flattened list. When used with HttpRequester, note the following behavior: - In HttpRequester.__post_init__, the entire list of backoff strategies is assigned to the error handler - However, the error handler's backoff_time() method only ever uses the first non-None strategy in the list - This means that if any backoff strategies are present, the first non-None strategy becomes the default - This applies to both user-defined response filters and errors from DEFAULT_ERROR_MAPPING - The list structure is not used to map different strategies to different error conditions - Therefore, subsequent strategies in the list will not be used Returns None if no handlers have strategies defined, which will result in HttpRequester using its default backoff strategy. + + Returns: + Optional[List[BackoffStrategy]]: A list of backoff strategies from all child handlers, + or None if no strategies are defined. """
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (1)
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py
(2 hunks)
⏰ Context from checks skipped due to timeout of 90000ms (4)
- GitHub Check: Pytest (Fast)
- GitHub Check: Pytest (All, Python 3.11, Ubuntu)
- GitHub Check: Pytest (All, Python 3.10, Ubuntu)
- GitHub Check: Analyze (python)
🔇 Additional comments (1)
airbyte_cdk/sources/declarative/requesters/error_handlers/composite_error_handler.py (1)
11-11
: LGTM! Clean import addition.The BackoffStrategy import is well-placed and properly scoped for the new functionality.
* remotes/airbyte/main: fix(airbyte-cdk): Fix RequestOptionsProvider for PerPartitionWithGlobalCursor (airbytehq#254) feat(low-code): add profile assertion flow to oauth authenticator component (airbytehq#236) feat(Low-Code Concurrent CDK): Add ConcurrentPerPartitionCursor (airbytehq#111) fix: don't mypy unit_tests (airbytehq#241) fix: handle backoff_strategies in CompositeErrorHandler (airbytehq#225) feat(concurrent cursor): attempt at clamping datetime (airbytehq#234) ci: use `ubuntu-24.04` explicitly (resolves CI warnings) (airbytehq#244) Fix(sdm): module ref issue in python components import (airbytehq#243) feat(source-declarative-manifest): add support for custom Python components from dynamic text input (airbytehq#174) chore(deps): bump avro from 1.11.3 to 1.12.0 (airbytehq#133) docs: comments on what the `Dockerfile` is for (airbytehq#240) chore: move ruff configuration to dedicated ruff.toml file (airbytehq#237)
Created a DPath Enhancing Extractor Refactored the record enhancement logic - moved to the extracted class Split the tests of DPathExtractor and DPathEnhancingExtractor Fix the failing tests: FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_custom_components[test_create_custom_component_with_subcomponent_that_uses_parameters] FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_custom_components_do_not_contain_extra_fields FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_parse_custom_component_fields_if_subcomponent FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_page_increment FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_offset_increment FAILED unit_tests/sources/file_based/test_file_based_scenarios.py::test_file_based_read[simple_unstructured_scenario] FAILED unit_tests/sources/file_based/test_file_based_scenarios.py::test_file_based_read[no_file_extension_unstructured_scenario] They faile because of comparing string and int values of the page_size (public) attribute. Imposed an invariant: on construction, page_size can be set to a string or int keep only values of one type in page_size for uniform comparison (convert the values of the other type) _page_size holds the internal / working value ... unless manipulated directly. Merged: feat(low-code concurrent): Allow async job low-code streams that are incremental to be run by the concurrent framework (airbytehq#228) fix(low-code): Fix declarative low-code state migration in SubstreamPartitionRouter (airbytehq#267) feat: combine slash command jobs into single job steps (airbytehq#266) feat(low-code): add items and property mappings to dynamic schemas (airbytehq#256) feat: add help response for unrecognized slash commands (airbytehq#264) ci: post direct links to html connector test reports (airbytehq#252) (airbytehq#263) fix(low-code): Fix legacy state migration in SubstreamPartitionRouter (airbytehq#261) fix(airbyte-cdk): Fix RequestOptionsProvider for PerPartitionWithGlobalCursor (airbytehq#254) feat(low-code): add profile assertion flow to oauth authenticator component (airbytehq#236) feat(Low-Code Concurrent CDK): Add ConcurrentPerPartitionCursor (airbytehq#111) fix: don't mypy unit_tests (airbytehq#241) fix: handle backoff_strategies in CompositeErrorHandler (airbytehq#225) feat(concurrent cursor): attempt at clamping datetime (airbytehq#234) fix(airbyte-cdk): Fix RequestOptionsProvider for PerPartitionWithGlobalCursor (airbytehq#254) feat(low-code): add profile assertion flow to oauth authenticator component (airbytehq#236) feat(Low-Code Concurrent CDK): Add ConcurrentPerPartitionCursor (airbytehq#111) fix: don't mypy unit_tests (airbytehq#241) fix: handle backoff_strategies in CompositeErrorHandler (airbytehq#225) feat(concurrent cursor): attempt at clamping datetime (airbytehq#234) ci: use `ubuntu-24.04` explicitly (resolves CI warnings) (airbytehq#244) Fix(sdm): module ref issue in python components import (airbytehq#243) feat(source-declarative-manifest): add support for custom Python components from dynamic text input (airbytehq#174) chore(deps): bump avro from 1.11.3 to 1.12.0 (airbytehq#133) docs: comments on what the `Dockerfile` is for (airbytehq#240) chore: move ruff configuration to dedicated ruff.toml file (airbytehq#237) Fix(sdm): module ref issue in python components import (airbytehq#243) feat(low-code): add DpathFlattenFields (airbytehq#227) feat(source-declarative-manifest): add support for custom Python components from dynamic text input (airbytehq#174) chore(deps): bump avro from 1.11.3 to 1.12.0 (airbytehq#133) docs: comments on what the `Dockerfile` is for (airbytehq#240) chore: move ruff configuration to dedicated ruff.toml file (airbytehq#237)
At record extraction step, in each record add the service field $root holding a reference to: * the root response object, when parsing JSON format * the original record, when parsing JSONL format that each record to process is extracted from. More service fields could be added in future. The service fields are available in the record's filtering and transform steps. Avoid: * reusing the maps/dictionaries produced, thus avoid building cyclic structures * transforming the service fields in the Flatten transformation. Explicitly cleanup the service field(s) after the transform step, thus making them: * local for the filter and transform steps * not visible to the next mapping and store steps (as they should be) * not visible in the tests beyond the test_record_selector (as they should be) This allows the record transformation logic to define its "local variables" to reuse some interim calculations. The contract of body parsing seems irregular in representing the cases of bad JSON, no JSON and empty JSON. Cannot be unified as that that irregularity is already used. Update the development environment setup documentation * to organize and present the setup steps explicitly * to avoid misunderstandings and wasted efforts. Update CONTRIBUTING.md to * collect and organize the knowledge on running the test locally. * state the actual testing steps. * clarify and make explicit the procedures and steps. The unit, integration, and acceptance tests in this exactly version succeed under Fedora 41, while one of them fails under Oracle Linux 8.7. not related to the contents of this PR. The integration tests of the CDK fail due to missing `secrets/config.json` file for the Shopify source. See airbytehq#197 Polish Integrate the DpathEnhancingExtractor in the UI of Airbyte. Created a DPath Enhancing Extractor Refactored the record enhancement logic - moved to the extracted class Split the tests of DPathExtractor and DPathEnhancingExtractor Fix the failing tests: FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_custom_components[test_create_custom_component_with_subcomponent_that_uses_parameters] FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_custom_components_do_not_contain_extra_fields FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_parse_custom_component_fields_if_subcomponent FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_page_increment FAILED unit_tests/sources/declarative/parsers/test_model_to_component_factory.py::test_create_offset_increment FAILED unit_tests/sources/file_based/test_file_based_scenarios.py::test_file_based_read[simple_unstructured_scenario] FAILED unit_tests/sources/file_based/test_file_based_scenarios.py::test_file_based_read[no_file_extension_unstructured_scenario] They faile because of comparing string and int values of the page_size (public) attribute. Imposed an invariant: on construction, page_size can be set to a string or int keep only values of one type in page_size for uniform comparison (convert the values of the other type) _page_size holds the internal / working value ... unless manipulated directly. Merged: feat(low-code concurrent): Allow async job low-code streams that are incremental to be run by the concurrent framework (airbytehq#228) fix(low-code): Fix declarative low-code state migration in SubstreamPartitionRouter (airbytehq#267) feat: combine slash command jobs into single job steps (airbytehq#266) feat(low-code): add items and property mappings to dynamic schemas (airbytehq#256) feat: add help response for unrecognized slash commands (airbytehq#264) ci: post direct links to html connector test reports (airbytehq#252) (airbytehq#263) fix(low-code): Fix legacy state migration in SubstreamPartitionRouter (airbytehq#261) fix(airbyte-cdk): Fix RequestOptionsProvider for PerPartitionWithGlobalCursor (airbytehq#254) feat(low-code): add profile assertion flow to oauth authenticator component (airbytehq#236) feat(Low-Code Concurrent CDK): Add ConcurrentPerPartitionCursor (airbytehq#111) fix: don't mypy unit_tests (airbytehq#241) fix: handle backoff_strategies in CompositeErrorHandler (airbytehq#225) feat(concurrent cursor): attempt at clamping datetime (airbytehq#234) fix(airbyte-cdk): Fix RequestOptionsProvider for PerPartitionWithGlobalCursor (airbytehq#254) feat(low-code): add profile assertion flow to oauth authenticator component (airbytehq#236) feat(Low-Code Concurrent CDK): Add ConcurrentPerPartitionCursor (airbytehq#111) fix: don't mypy unit_tests (airbytehq#241) fix: handle backoff_strategies in CompositeErrorHandler (airbytehq#225) feat(concurrent cursor): attempt at clamping datetime (airbytehq#234) ci: use `ubuntu-24.04` explicitly (resolves CI warnings) (airbytehq#244) Fix(sdm): module ref issue in python components import (airbytehq#243) feat(source-declarative-manifest): add support for custom Python components from dynamic text input (airbytehq#174) chore(deps): bump avro from 1.11.3 to 1.12.0 (airbytehq#133) docs: comments on what the `Dockerfile` is for (airbytehq#240) chore: move ruff configuration to dedicated ruff.toml file (airbytehq#237) Fix(sdm): module ref issue in python components import (airbytehq#243) feat(low-code): add DpathFlattenFields (airbytehq#227) feat(source-declarative-manifest): add support for custom Python components from dynamic text input (airbytehq#174) chore(deps): bump avro from 1.11.3 to 1.12.0 (airbytehq#133) docs: comments on what the `Dockerfile` is for (airbytehq#240) chore: move ruff configuration to dedicated ruff.toml file (airbytehq#237) formatted Update record_extractor.py Trigger a new build. Hopefully, the integration test infrastructure is fixed. Update CONTRIBUTING.md Trigger a new build
What
Although users can define custom backoff logic in the manifest, the CompositeErrorHandler component does not currently access the backoff_strategies, meaning it always defaults to the default exponential behavior. By accessing the
backoff_strategies
in the childErrorHandler
components, we can handle user-specified strategies.Resolves: airbytehq/airbyte#42928
Cautionary Note
My understanding of the ErrorHandler flow is that it is not currently configured to respect multiple backoff strategies for a single stream, and will always use the first non-
None
value as the strategy for both:response_filters
For example, using:
We would expect a
500
level response to utilize the Exponential strategy. However, the output seen is in fact:This is a function of the errorhandler logic flow and is a known behavior, distinct from the bug this PR tries to address.
Summary by CodeRabbit
New Features
Tests