[data] Disable DataSourceV2 by default#63674
Merged
goutamvenkat-anyscale merged 2 commits intoJun 1, 2026
Merged
Conversation
![gemini-code-assist[bot]](https://cdn.statically.io/img/avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
goutamvenkat-anyscale
force-pushed
the
goutam/data-disable-datasource-v2-default
branch
from
457c636 to
e95b6c8
Compare
V2 currently OOMs on shuffle-heavy workloads because ``ReadFiles.infer_metadata()`` returns no ``size_bytes``, dropping ``HashShufflingOperatorBase._get_default_aggregator_ray_remote_args`` into its 1-GiB-per-aggregator fallback (e.g., TPC-H Q9 SF=100 autoscaling release test). The fix is on a separate branch (sample-extrapolated estimate surfaced via ``ReadFiles.infer_metadata``); flip the default off until that lands and the release tests pass. Users can still opt in with ``DataContext.use_datasource_v2 = True``. Signed-off-by: Goutam <goutam@anyscale.com>
goutamvenkat-anyscale
force-pushed
the
goutam/data-disable-datasource-v2-default
branch
from
63dbe87 to
ff7b958
Compare
goutamvenkat-anyscale
deleted the
goutam/data-disable-datasource-v2-default
branch
rueian
pushed a commit
to rueian/ray
that referenced
this pull request
Jun 4, 2026
## Summary Flip ``DEFAULT_USE_DATASOURCE_V2`` from ``True`` to ``False``. V2 currently OOMs on shuffle-heavy workloads because ``ReadFiles.infer_metadata()`` returns no ``size_bytes``, dropping ``HashShufflingOperatorBase._get_default_aggregator_ray_remote_args`` into its 1-GiB-per-aggregator fallback. Most visibly: TPC-H Q9 SF=100 autoscaling release test fails with ``HashShuffleAggregator`` actor death from host-level OOM. The fix (sample-extrapolated ``size_bytes`` surfaced via ``ReadFiles.infer_metadata``) is on a separate branch — flipping this off keeps master green until that lands and the release tests pass. Users can still opt in via ``DataContext.use_datasource_v2 = True``. ## Test plan - [ ] CI is green with V1 default (no V2 regressions reintroduced). - [ ] Run TPC-H Q9 autoscaling release test on master with this flip — expect pass. - [ ] Confirm ``read_parquet`` users on V1 path see unchanged behavior. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Signed-off-by: Goutam <goutam@anyscale.com>
limarkdcunha
pushed a commit
to limarkdcunha/ray
that referenced
this pull request
Jun 30, 2026
## Summary Flip ``DEFAULT_USE_DATASOURCE_V2`` from ``True`` to ``False``. V2 currently OOMs on shuffle-heavy workloads because ``ReadFiles.infer_metadata()`` returns no ``size_bytes``, dropping ``HashShufflingOperatorBase._get_default_aggregator_ray_remote_args`` into its 1-GiB-per-aggregator fallback. Most visibly: TPC-H Q9 SF=100 autoscaling release test fails with ``HashShuffleAggregator`` actor death from host-level OOM. The fix (sample-extrapolated ``size_bytes`` surfaced via ``ReadFiles.infer_metadata``) is on a separate branch — flipping this off keeps master green until that lands and the release tests pass. Users can still opt in via ``DataContext.use_datasource_v2 = True``. ## Test plan - [ ] CI is green with V1 default (no V2 regressions reintroduced). - [ ] Run TPC-H Q9 autoscaling release test on master with this flip — expect pass. - [ ] Confirm ``read_parquet`` users on V1 path see unchanged behavior. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Signed-off-by: Goutam <goutam@anyscale.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Flip
DEFAULT_USE_DATASOURCE_V2fromTruetoFalse. V2 currently OOMs on shuffle-heavy workloads becauseReadFiles.infer_metadata()returns nosize_bytes, droppingHashShufflingOperatorBase._get_default_aggregator_ray_remote_argsinto its 1-GiB-per-aggregator fallback. Most visibly: TPC-H Q9 SF=100 autoscaling release test fails withHashShuffleAggregatoractor death from host-level OOM.The fix (sample-extrapolated
size_bytessurfaced viaReadFiles.infer_metadata) is on a separate branch — flipping this off keeps master green until that lands and the release tests pass.Users can still opt in via
DataContext.use_datasource_v2 = True.Test plan
read_parquetusers on V1 path see unchanged behavior.🤖 Generated with Claude Code