Fix start_trace compatibility with old mlflow clients#19183
Fix start_trace compatibility with old mlflow clients#19183B-Step62 merged 7 commits intomlflow:masterfrom
Conversation
Signed-off-by: Serena Ruan <serena.rxy@gmail.com>
|
Documentation preview for 5a1ce25 is available at: More info
|
|
|
||
| @pytest.mark.skipif( | ||
| mlflow.get_tracking_uri().startswith("mysql"), | ||
| reason="MySQL does not support concurrent log_spans calls for now", |
There was a problem hiding this comment.
This test fails on mysql with below error:
sqlalchemy.exc.OperationalError: (MySQLdb.OperationalError) (1213, 'Deadlock found when trying to get lock; try restarting transaction')
[SQL: UPDATE trace_info SET timestamp_ms=CASE WHEN (trace_info.timestamp_ms > %s) THEN %s ELSE trace_info.timestamp_ms END, execution_time_ms=(CASE WHEN (trace_info.timestamp_ms + trace_info.execution_time_ms > %s) THEN trace_info.timestamp_ms + trace_info.execution_time_ms ELSE %s END - CASE WHEN (trace_info.timestamp_ms > %s) THEN %s ELSE trace_info.timestamp_ms END) WHERE trace_info.request_id = %s]
[parameters: (1000, 1000, 2000, 2000, 1000, 1000, 'tr-46617ca3256749c7a5859266ad843913')]
But it's not related to the current change, but because of session.query(SqlTraceInfo).filter(SqlTraceInfo.request_id == trace_id).update call. I'm skipping this test in mysql for now, but we need to fix this issue separately.
Repro command:
uv run docker compose --project-directory tests/db run --rm --no-TTY mlflow-mysql pytest tests/store/tracking/test_sqlalchemy_store.py::test_concurrent_log_spans_spans_location_tag
There was a problem hiding this comment.
Do we plan to fix this by 3.7?
There was a problem hiding this comment.
Pull request overview
This PR fixes a compatibility issue between newer MLflow servers and older MLflow clients for trace management. The issue occurred because old clients store traces in artifact storage but still invoke start_trace, while the new server was automatically adding a SPANS_LOCATION tag, causing traces to be incorrectly loaded from the span table instead of artifact storage.
Key changes:
- Remove automatic
SPANS_LOCATIONtag creation instart_trace()to support old clients that store spans in artifact storage - Add logic to preserve the
SPANS_LOCATIONtag when it's set bylog_spans()for new clients - Update
log_spans()to set theSPANS_LOCATIONtag when spans are stored in the tracking database
Reviewed changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
mlflow/store/tracking/sqlalchemy_store.py |
Removed automatic SPANS_LOCATION tag in start_trace() and added logic to preserve it if set by log_spans(), plus added tag update in log_spans() |
tests/store/tracking/test_sqlalchemy_store.py |
Updated existing tests to remove SPANS_LOCATION tag expectations and added new tests for different call order scenarios |
libs/typescript/core/tests/clients/client.test.ts |
Removed expectation of SPANS_LOCATION tag in TypeScript test |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Signed-off-by: Serena Ruan <serena.rxy@gmail.com>
Signed-off-by: Serena Ruan <serena.rxy@gmail.com>
🛠 DevTools 🛠
Install mlflow from this PR
For Databricks, use the following command:
Related Issues/PRs
#xxxWhat changes are proposed in this pull request?
In old mlflow clients traces are stored in artifact storage, but it still invokes 'start_trace' call. However the SPANS_LOCATION tag is added, which decides where to load the spans. As a result, this causes an issue with newer mlflow server + old mlflow clients, where traces are stored in artifact repository but trying to be loaded from span table.
How is this PR tested?
Does this PR require documentation update?
Release Notes
Is this a user-facing change?
What component(s), interfaces, languages, and integrations does this PR affect?
Components
area/tracking: Tracking Service, tracking client APIs, autologgingarea/models: MLmodel format, model serialization/deserialization, flavorsarea/model-registry: Model Registry service, APIs, and the fluent client calls for Model Registryarea/scoring: MLflow Model server, model deployment tools, Spark UDFsarea/evaluation: MLflow model evaluation features, evaluation metrics, and evaluation workflowsarea/gateway: MLflow AI Gateway client APIs, server, and third-party integrationsarea/prompts: MLflow prompt engineering features, prompt templates, and prompt managementarea/tracing: MLflow Tracing features, tracing APIs, and LLM tracing functionalityarea/projects: MLproject format, project running backendsarea/uiux: Front-end, user experience, plotting, JavaScript, JavaScript dev serverarea/build: Build and test infrastructure for MLflowarea/docs: MLflow documentation pagesHow should the PR be classified in the release notes? Choose one:
rn/none- No description will be included. The PR will be mentioned only by the PR number in the "Small Bugfixes and Documentation Updates" sectionrn/breaking-change- The PR will be mentioned in the "Breaking Changes" sectionrn/feature- A new user-facing feature worth mentioning in the release notesrn/bug-fix- A user-facing bug fix worth mentioning in the release notesrn/documentation- A user-facing documentation change worth mentioning in the release notesShould this PR be included in the next patch release?
Yesshould be selected for bug fixes, documentation updates, and other small changes.Noshould be selected for new features and larger changes. If you're unsure about the release classification of this PR, leave this unchecked to let the maintainers decide.What is a minor/patch release?
Bug fixes, doc updates and new features usually go into minor releases.
Bug fixes and doc updates usually go into patch releases.