Set default connection timeout

[harupy]

🛠 DevTools 🛠

Install mlflow from this PR

pip install git+https://github.com/mlflow/mlflow.git@refs/pull/12016/merge

Checkout with GitHub CLI

gh pr checkout 12016

Related Issues/PRs

What changes are proposed in this pull request?

Address #11969 (comment)

How is this PR tested?

• Existing unit/integration tests
• New unit/integration tests
• Manual tests

Does this PR require documentation update?

• No. You can skip the rest of this section.
• Yes. I've updated:
  • Examples
  • API references
  • Instructions

Release Notes

Is this a user-facing change?

• No. You can skip the rest of this section.
• Yes. Give a description of this change to be included in the release notes for MLflow users.

What component(s), interfaces, languages, and integrations does this PR affect?

Components

• area/artifacts: Artifact stores and artifact logging
• area/build: Build and test infrastructure for MLflow
• area/deployments: MLflow Deployments client APIs, server, and third-party Deployments integrations
• area/docs: MLflow documentation pages
• area/examples: Example code
• area/model-registry: Model Registry service, APIs, and the fluent client calls for Model Registry
• area/models: MLmodel format, model serialization/deserialization, flavors
• area/recipes: Recipes, Recipe APIs, Recipe configs, Recipe Templates
• area/projects: MLproject format, project running backends
• area/scoring: MLflow Model server, model deployment tools, Spark UDFs
• area/server-infra: MLflow Tracking server backend
• area/tracking: Tracking Service, tracking client APIs, autologging

Interface

• area/uiux: Front-end, user experience, plotting, JavaScript, JavaScript dev server
• area/docker: Docker use across MLflow's components, such as MLflow Projects and MLflow Models
• area/sqlalchemy: Use of SQLAlchemy in the Tracking Service or Model Registry
• area/windows: Windows support

Language

• language/r: R APIs and clients
• language/java: Java APIs and clients
• language/new: Proposals for new client languages

Integrations

• integrations/azure: Azure and Azure ML integrations
• integrations/sagemaker: SageMaker integrations
• integrations/databricks: Databricks integrations

How 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" section
• rn/breaking-change - The PR will be mentioned in the "Breaking Changes" section
• rn/feature - A new user-facing feature worth mentioning in the release notes
• rn/bug-fix - A user-facing bug fix worth mentioning in the release notes
• rn/documentation - A user-facing documentation change worth mentioning in the release notes

Should this PR be included in the next patch release?

Yes should be selected for bug fixes, documentation updates, and other small changes. No should 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?

• Minor release: a release that increments the second part of the version number (e.g., 1.2.0 -> 1.3.0).
  Bug fixes, doc updates and new features usually go into minor releases.
• Patch release: a release that increments the third part of the version number (e.g., 1.2.0 -> 1.2.1).
  Bug fixes and doc updates usually go into patch releases.

• Yes (this PR will be cherry-picked and included in the next patch release)
• No (this PR will be included in the next minor release)

[github-actions]

Documentation preview for 47bd07f will be available when this CircleCI job
completes successfully.

• Top page
• Changed pages

More info

• Ignore this comment if this PR does not change the documentation.
• It takes a few minutes for the preview to be available.
• The preview is updated when a new commit is pushed to this PR.
• This comment was created by https://github.com/mlflow/mlflow/actions/runs/9104530791.

[harupy]

Suggested change

	read request. Default to (60, None) which means 60 seconds for connect and no timeout
	for read.
	read request. Default to (30, None) which means 30 seconds for connect and no timeout
	for read.

[BenWilson2]

This seems reasonable. Let's test with a massive model using MPU (Mixtral) on s3 and ADLS to verify.

[dbczumar]

@harupy A couple questions:

1. Are we sure this works? @smurching informed that @dakinggg tried something similar, but it didn't resolve the issue that he was facing with ADLS during fine tuning. (@dakinggg can you comment further here?)

2. What's our retry policy when there's a connection timeout? We need to make sure we're retrying :)

[harupy]

@dbczumar

1. No, we aren't. We need to test this. We can close this PR if we've already tried this and it didn't work :)
2. Can we do retries retry with exponential backoff on requests.exceptions.ConnectTimeout?

for _ in range(max_attempts):
    try:
        ...
    except requests.exceptions.ConnectTimeout:
        ...

[dbczumar]

1. Can you work with @dakinggg to test this further / explore alternatives if this doesn't work?
2. Absolutely - we must do something like this

[harupy]

@dbczumar We already using retry logic for connection related errors:

mlflow/mlflow/utils/request_utils.py

Line 120 in 125402c

"connect": max_retries,

[dakinggg]

@harupy @dbczumar

Unfortunately reproducing is quite difficult (we had to run many training runs in order to repro, and never managed to get a standalone uploading script to repro). It also seemed to interact with load (i.e. was possibly worse during the day). I think we were hoping that as part of mitigating this on the mlflow side, we could carry on @smurching attempt to reproduce in a standalone script.

That being said, it did seem that adding the explicit request.close() (mosaicml/composer#3276) was important. The request timeout in this PR was not verified to work on its own (although seems like a good idea). The socket.setdefaulttimeout from mosaicml/composer#3265 was also not verified to work on its own, however this one may be a little dangerous since its a global default. Not entirely sure there.