feat: Added batching to feature server /push to offline store (#5683) by jfw-ppi · Pull Request #5729 · feast-dev/feast

jfw-ppi · 2025-11-18T18:44:39Z

What this PR does / why we need it:

Added batching configuration for feature_servers /push endpoint for offline store writes

Which issue(s) this PR fixes:

franciscojavierarceo · 2025-11-21T04:18:00Z

sdk/python/feast/infra/feature_servers/base_config.py


 from feast.repo_config import FeastConfigBaseModel

+class OfflinePushBatchingConfig(FeastConfigBaseModel):


i think having a config is fine but do we actually need it? we probably could have just passed these as optional args, no?

You're right, I just noticed the FeatureLoggingConfig with 5 fields and decided that for 3 fields it would be justified to also add a config. Do you want me to refactor it to use optional args?

i think having a config is fine but do we actually need it? we probably could have just passed these as optional args, no?

I refactored it as you wanted, so that there is no config. @franciscojavierarceo

…dev#5683) Signed-off-by: Jacob Weinhold <[email protected]> fix: formatting,l int errors (feast-dev#5683) Signed-off-by: Jacob Weinhold <[email protected]>

…push

…push Signed-off-by: Jacob Weinhold <[email protected]>

ntkathole · 2025-12-31T05:17:50Z

sdk/python/feast/feature_server.py

+            return
+
+        batch_df = pd.concat(dfs, ignore_index=True)
+        self._buffers[key].clear()


Does it make sense to move clear inside try: self._store.push so that buffer gets cleared only after the write succeeds ?

Totally makes sense, it's done. Thanks for seeing that.

ntkathole · 2025-12-31T05:25:13Z

sdk/python/feast/feature_server.py

+
+        # NOTE: offline writes are currently synchronous only, so we call directly
+        try:
+            self._store.push(


What about splitting _flush_locked into two methods: one that extracts data (with lock) and one that does I/O (without lock) ?

Something like:

Extracting the batch data while holding the lock

Releasing the lock before doing I/O

Re-enqueueing data if the write fails

Thanks for the suggestion!
I split _flush_locked into _drain_locked (extract under lock) and _flush (I/O without lock). I also added _inflight to prevent concurrent flushes per key. On failure the drained batch is re‑enqueued so we don’t drop data.

ntkathole · 2025-12-31T05:31:17Z

sdk/python/tests/unit/test_feature_server.py

+
+    # use a multi-row payload to ensure we test non-trivial dfs
+    resp = client.post("/push", json=push_body_many(push_mode, count=2, id_start=100))
+    assert resp.status_code == 200


optional but I think it's good to return 202 when batching is enabled and offline writes are involved

Nice catch! It's done.

Copilot

Pull request overview

This PR adds configurable batching support for offline writes to the feature server's /push endpoint. The batching mechanism buffers offline writes and flushes them based on either a size threshold or time interval, improving throughput for high-volume offline push operations.

Key Changes:

Introduced OfflineWriteBatcher class that manages buffered writes in a background thread
Added configuration options for batch size and interval in BaseFeatureServerConfig
Modified /push endpoint logic to separate online and offline writes when batching is enabled

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
sdk/python/feast/feature_server.py	Implemented `OfflineWriteBatcher` class and integrated batching logic into the `/push` endpoint
sdk/python/feast/infra/feature_servers/base_config.py	Added three new configuration fields for offline push batching
sdk/python/tests/unit/test_feature_server.py	Added comprehensive test coverage for batching behavior across different push modes and configurations
docs/reference/feature-store-yaml.md	Documented the new `feature_server` configuration block with batching options
docs/reference/feature-servers/python-feature-server.md	Added user-facing documentation explaining offline write batching functionality

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2025-12-31T11:10:27Z

sdk/python/feast/feature_server.py

+            allow_registry_cache = request.allow_registry_cache
+            transform_on_write = request.transform_on_write
+
+            # Async currently only applies to online store writes (ONLINE / ONLINE_AND_OFFLINE paths) as theres no async for offline store


Corrected spelling: 'theres' should be 'there's'.

Suggested change

# Async currently only applies to online store writes (ONLINE / ONLINE_AND_OFFLINE paths) as theres no async for offline store

# Async currently only applies to online store writes (ONLINE / ONLINE_AND_OFFLINE paths) as there's no async for offline store

Copilot · 2025-12-31T11:10:27Z

sdk/python/tests/unit/test_feature_server.py

+    fs, enabled: bool = True, batch_size: int = 1, batch_interval_seconds: int = 60
+):
+    """
+        Attach a minimal feature_server.offline_push_batching config


The docstring has inconsistent indentation. The closing triple quotes and the description line should be aligned with the opening triple quotes for standard formatting.

Suggested change

Attach a minimal feature_server.offline_push_batching config

Attach a minimal feature_server.offline_push_batching config

…-dev#5683](feast-dev#5683)) Signed-off-by: Jacob Weinhold <[email protected]>

…push

ntkathole

lgtm

…push

# [0.59.0](v0.58.0...v0.59.0) (2026-01-16) ### Bug Fixes * Add get_table_query_string_with_alias() for PostgreSQL subquery aliasing ([#5811](#5811)) ([11122ce](11122ce)) * Add hybrid online store to ONLINE_STORE_CLASS_FOR_TYPE mapping ([#5810](#5810)) ([678589b](678589b)) * Add possibility to overwrite send_receive_timeout for clickhouse offline store ([#5792](#5792)) ([59dbb33](59dbb33)) * Denial by default to all resources when no permissions set ([#5663](#5663)) ([1524f1c](1524f1c)) * Make operator include full OIDC secret in repo config ([#5676](#5676)) ([#5809](#5809)) ([a536bc2](a536bc2)) * Populate Postgres `registry.path` during `feast init` ([#5785](#5785)) ([f293ae8](f293ae8)) * **redis:** Preserve millisecond timestamp precision for Redis online store ([#5807](#5807)) ([9e3f213](9e3f213)) * Search API to return all matching tags in matched_tags field ([#5843](#5843)) ([de37f66](de37f66)) * Spark Materialization Engine Cannot Infer Schema ([#5806](#5806)) ([58d0325](58d0325)), closes [#5594](#5594) [#5594](#5594) * Support arro3 table schema with newer deltalake packages ([#5799](#5799)) ([103c5e9](103c5e9)) * Timestamp formatting and lakehouse-type connector for trino_offline_store. ([#5846](#5846)) ([c2ea7e9](c2ea7e9)) * Update model_validator to use instance method signature (Pydantic v2.12 deprecation) ([#5825](#5825)) ([3c10b6e](3c10b6e)) ### Features * Add dbt integration for importing models as FeatureViews ([#5827](#5827)) ([b997361](b997361)), closes [#3335](#3335) [#3335](#3335) [#3335](#3335) * Add GCS registry store in Go feature server ([#5818](#5818)) ([1dc2be5](1dc2be5)) * Add progress bar to CLI from feast apply ([#5867](#5867)) ([ab3562b](ab3562b)) * Add RBAC blog post to website ([#5861](#5861)) ([b1844a3](b1844a3)) * Add skip_feature_view_validation parameter to FeatureStore.apply() and plan() ([#5859](#5859)) ([5482a0e](5482a0e)) * Added batching to feature server /push to offline store ([#5683](#5683)) ([#5729](#5729)) ([ce35ce6](ce35ce6)) * Enable static artifacts for feature server that can be used in Feature Transformations ([#5787](#5787)) ([edefc3f](edefc3f)) * Improve lambda materialization engine ([#5829](#5829)) ([f6116f9](f6116f9)) * Offline Store historical features retrieval based on datetime range in Ray ([#5738](#5738)) ([e484c12](e484c12)) * Read, Save docs and chat fixes ([#5865](#5865)) ([2081b55](2081b55)) * Resolve pyarrow >21 installation with ibis-framework ([#5847](#5847)) ([8b9bb50](8b9bb50)) * Support staging for spark materialization ([#5671](#5671)) ([#5797](#5797)) ([5b787af](5b787af))

# [0.59.0](feast-dev/feast@v0.58.0...v0.59.0) (2026-01-16) ### Bug Fixes * Add get_table_query_string_with_alias() for PostgreSQL subquery aliasing ([feast-dev#5811](feast-dev#5811)) ([11122ce](feast-dev@11122ce)) * Add hybrid online store to ONLINE_STORE_CLASS_FOR_TYPE mapping ([feast-dev#5810](feast-dev#5810)) ([678589b](feast-dev@678589b)) * Add possibility to overwrite send_receive_timeout for clickhouse offline store ([feast-dev#5792](feast-dev#5792)) ([59dbb33](feast-dev@59dbb33)) * Denial by default to all resources when no permissions set ([feast-dev#5663](feast-dev#5663)) ([1524f1c](feast-dev@1524f1c)) * Make operator include full OIDC secret in repo config ([feast-dev#5676](feast-dev#5676)) ([feast-dev#5809](feast-dev#5809)) ([a536bc2](feast-dev@a536bc2)) * Populate Postgres `registry.path` during `feast init` ([feast-dev#5785](feast-dev#5785)) ([f293ae8](feast-dev@f293ae8)) * **redis:** Preserve millisecond timestamp precision for Redis online store ([feast-dev#5807](feast-dev#5807)) ([9e3f213](feast-dev@9e3f213)) * Search API to return all matching tags in matched_tags field ([feast-dev#5843](feast-dev#5843)) ([de37f66](feast-dev@de37f66)) * Spark Materialization Engine Cannot Infer Schema ([feast-dev#5806](feast-dev#5806)) ([58d0325](feast-dev@58d0325)), closes [feast-dev#5594](feast-dev#5594) [feast-dev#5594](feast-dev#5594) * Support arro3 table schema with newer deltalake packages ([feast-dev#5799](feast-dev#5799)) ([103c5e9](feast-dev@103c5e9)) * Timestamp formatting and lakehouse-type connector for trino_offline_store. ([feast-dev#5846](feast-dev#5846)) ([c2ea7e9](feast-dev@c2ea7e9)) * Update model_validator to use instance method signature (Pydantic v2.12 deprecation) ([feast-dev#5825](feast-dev#5825)) ([3c10b6e](feast-dev@3c10b6e)) ### Features * Add dbt integration for importing models as FeatureViews ([feast-dev#5827](feast-dev#5827)) ([b997361](feast-dev@b997361)), closes [feast-dev#3335](feast-dev#3335) [feast-dev#3335](feast-dev#3335) [feast-dev#3335](feast-dev#3335) * Add GCS registry store in Go feature server ([feast-dev#5818](feast-dev#5818)) ([1dc2be5](feast-dev@1dc2be5)) * Add progress bar to CLI from feast apply ([feast-dev#5867](feast-dev#5867)) ([ab3562b](feast-dev@ab3562b)) * Add RBAC blog post to website ([feast-dev#5861](feast-dev#5861)) ([b1844a3](feast-dev@b1844a3)) * Add skip_feature_view_validation parameter to FeatureStore.apply() and plan() ([feast-dev#5859](feast-dev#5859)) ([5482a0e](feast-dev@5482a0e)) * Added batching to feature server /push to offline store ([feast-dev#5683](feast-dev#5683)) ([feast-dev#5729](feast-dev#5729)) ([ce35ce6](feast-dev@ce35ce6)) * Enable static artifacts for feature server that can be used in Feature Transformations ([feast-dev#5787](feast-dev#5787)) ([edefc3f](feast-dev@edefc3f)) * Improve lambda materialization engine ([feast-dev#5829](feast-dev#5829)) ([f6116f9](feast-dev@f6116f9)) * Offline Store historical features retrieval based on datetime range in Ray ([feast-dev#5738](feast-dev#5738)) ([e484c12](feast-dev@e484c12)) * Read, Save docs and chat fixes ([feast-dev#5865](feast-dev#5865)) ([2081b55](feast-dev@2081b55)) * Resolve pyarrow >21 installation with ibis-framework ([feast-dev#5847](feast-dev#5847)) ([8b9bb50](feast-dev@8b9bb50)) * Support staging for spark materialization ([feast-dev#5671](feast-dev#5671)) ([feast-dev#5797](feast-dev#5797)) ([5b787af](feast-dev@5b787af)) Signed-off-by: Kyle DePasquale <[email protected]>

# [0.59.0](feast-dev/feast@v0.58.0...v0.59.0) (2026-01-16) ### Bug Fixes * Add get_table_query_string_with_alias() for PostgreSQL subquery aliasing ([feast-dev#5811](feast-dev#5811)) ([11122ce](feast-dev@11122ce)) * Add hybrid online store to ONLINE_STORE_CLASS_FOR_TYPE mapping ([feast-dev#5810](feast-dev#5810)) ([678589b](feast-dev@678589b)) * Add possibility to overwrite send_receive_timeout for clickhouse offline store ([feast-dev#5792](feast-dev#5792)) ([59dbb33](feast-dev@59dbb33)) * Denial by default to all resources when no permissions set ([feast-dev#5663](feast-dev#5663)) ([1524f1c](feast-dev@1524f1c)) * Make operator include full OIDC secret in repo config ([feast-dev#5676](feast-dev#5676)) ([feast-dev#5809](feast-dev#5809)) ([a536bc2](feast-dev@a536bc2)) * Populate Postgres `registry.path` during `feast init` ([feast-dev#5785](feast-dev#5785)) ([f293ae8](feast-dev@f293ae8)) * **redis:** Preserve millisecond timestamp precision for Redis online store ([feast-dev#5807](feast-dev#5807)) ([9e3f213](feast-dev@9e3f213)) * Search API to return all matching tags in matched_tags field ([feast-dev#5843](feast-dev#5843)) ([de37f66](feast-dev@de37f66)) * Spark Materialization Engine Cannot Infer Schema ([feast-dev#5806](feast-dev#5806)) ([58d0325](feast-dev@58d0325)), closes [feast-dev#5594](feast-dev#5594) [feast-dev#5594](feast-dev#5594) * Support arro3 table schema with newer deltalake packages ([feast-dev#5799](feast-dev#5799)) ([103c5e9](feast-dev@103c5e9)) * Timestamp formatting and lakehouse-type connector for trino_offline_store. ([feast-dev#5846](feast-dev#5846)) ([c2ea7e9](feast-dev@c2ea7e9)) * Update model_validator to use instance method signature (Pydantic v2.12 deprecation) ([feast-dev#5825](feast-dev#5825)) ([3c10b6e](feast-dev@3c10b6e)) ### Features * Add dbt integration for importing models as FeatureViews ([feast-dev#5827](feast-dev#5827)) ([b997361](feast-dev@b997361)), closes [feast-dev#3335](feast-dev#3335) [feast-dev#3335](feast-dev#3335) [feast-dev#3335](feast-dev#3335) * Add GCS registry store in Go feature server ([feast-dev#5818](feast-dev#5818)) ([1dc2be5](feast-dev@1dc2be5)) * Add progress bar to CLI from feast apply ([feast-dev#5867](feast-dev#5867)) ([ab3562b](feast-dev@ab3562b)) * Add RBAC blog post to website ([feast-dev#5861](feast-dev#5861)) ([b1844a3](feast-dev@b1844a3)) * Add skip_feature_view_validation parameter to FeatureStore.apply() and plan() ([feast-dev#5859](feast-dev#5859)) ([5482a0e](feast-dev@5482a0e)) * Added batching to feature server /push to offline store ([feast-dev#5683](feast-dev#5683)) ([feast-dev#5729](feast-dev#5729)) ([ce35ce6](feast-dev@ce35ce6)) * Enable static artifacts for feature server that can be used in Feature Transformations ([feast-dev#5787](feast-dev#5787)) ([edefc3f](feast-dev@edefc3f)) * Improve lambda materialization engine ([feast-dev#5829](feast-dev#5829)) ([f6116f9](feast-dev@f6116f9)) * Offline Store historical features retrieval based on datetime range in Ray ([feast-dev#5738](feast-dev#5738)) ([e484c12](feast-dev@e484c12)) * Read, Save docs and chat fixes ([feast-dev#5865](feast-dev#5865)) ([2081b55](feast-dev@2081b55)) * Resolve pyarrow >21 installation with ibis-framework ([feast-dev#5847](feast-dev#5847)) ([8b9bb50](feast-dev@8b9bb50)) * Support staging for spark materialization ([feast-dev#5671](feast-dev#5671)) ([feast-dev#5797](feast-dev#5797)) ([5b787af](feast-dev@5b787af)) Signed-off-by: yassinnouh21 <[email protected]>

jfw-ppi requested a review from a team as a code owner November 18, 2025 18:44

jfw-ppi mentioned this pull request Nov 18, 2025

Feature Request: Add configurable batching for offline store writes in Feature Server push API #5683

Closed

ntkathole changed the title ~~feat: added batching to feature server /push to offline store (#5683)~~ feat: Added batching to feature server /push to offline store (#5683) Nov 19, 2025

franciscojavierarceo reviewed Nov 21, 2025

View reviewed changes

jfw-ppi force-pushed the 5683-configurable-batching-feature-server-push branch 5 times, most recently from 494912d to 107bf09 Compare November 26, 2025 21:42

franciscojavierarceo added the ok-to-test label Nov 28, 2025

jfw-ppi force-pushed the 5683-configurable-batching-feature-server-push branch 2 times, most recently from c980db9 to 04dc34f Compare November 30, 2025 11:32

ntkathole requested a review from franciscojavierarceo December 15, 2025 16:07

feat: added batching to feature server /push to offline store (feast-…

1a3ccbd

…dev#5683) Signed-off-by: Jacob Weinhold <[email protected]> fix: formatting,l int errors (feast-dev#5683) Signed-off-by: Jacob Weinhold <[email protected]>

ntkathole force-pushed the 5683-configurable-batching-feature-server-push branch from 20079eb to 1a3ccbd Compare December 15, 2025 17:00

jfw-ppi added 3 commits December 18, 2025 20:14

Merge branch 'master' into 5683-configurable-batching-feature-server-…

49de0cb

…push

Merge branch 'master' into 5683-configurable-batching-feature-server-…

fbcc467

…push

Merge branch 'master' into 5683-configurable-batching-feature-server-…

bb299d9

…push Signed-off-by: Jacob Weinhold <[email protected]>

jfw-ppi force-pushed the 5683-configurable-batching-feature-server-push branch from a60c605 to bb299d9 Compare December 28, 2025 21:19

ntkathole reviewed Dec 31, 2025

View reviewed changes

franciscojavierarceo requested review from Copilot and removed request for franciscojavierarceo December 31, 2025 11:10

Copilot AI reviewed Dec 31, 2025

View reviewed changes

feat: added batching to feature server /push to offline store ([feast…

2747405

…-dev#5683](feast-dev#5683)) Signed-off-by: Jacob Weinhold <[email protected]>

jfw-ppi force-pushed the 5683-configurable-batching-feature-server-push branch from a99a9b0 to 2747405 Compare January 1, 2026 15:28

Merge branch 'master' into 5683-configurable-batching-feature-server-…

02899c1

…push

ntkathole approved these changes Jan 2, 2026

View reviewed changes

ntkathole requested a review from franciscojavierarceo January 2, 2026 08:13

Merge branch 'master' into 5683-configurable-batching-feature-server-…

a65937a

…push

franciscojavierarceo approved these changes Jan 3, 2026

View reviewed changes

franciscojavierarceo merged commit ce35ce6 into feast-dev:master Jan 3, 2026
19 checks passed


		from feast.repo_config import FeastConfigBaseModel

		class OfflinePushBatchingConfig(FeastConfigBaseModel):

	# Async currently only applies to online store writes (ONLINE / ONLINE_AND_OFFLINE paths) as theres no async for offline store
	# Async currently only applies to online store writes (ONLINE / ONLINE_AND_OFFLINE paths) as there's no async for offline store

	Attach a minimal feature_server.offline_push_batching config
	Attach a minimal feature_server.offline_push_batching config

Conversation

jfw-ppi commented Nov 18, 2025

What this PR does / why we need it:

Which issue(s) this PR fixes:

Uh oh!

franciscojavierarceo Nov 21, 2025

Choose a reason for hiding this comment

Uh oh!

jfw-ppi Nov 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jfw-ppi Nov 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ntkathole Dec 31, 2025

Choose a reason for hiding this comment

Uh oh!

jfw-ppi Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

ntkathole Dec 31, 2025

Choose a reason for hiding this comment

Uh oh!

jfw-ppi Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

ntkathole Dec 31, 2025

Choose a reason for hiding this comment

Uh oh!

jfw-ppi Jan 1, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Dec 31, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 31, 2025

Choose a reason for hiding this comment

Uh oh!

ntkathole left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

jfw-ppi Nov 23, 2025 •

edited

Loading

jfw-ppi Nov 26, 2025 •

edited

Loading