[Data] Removing Parquet metadata fetching in ParquetDatasource
#56105
Conversation
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
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request effectively removes the inefficient Parquet metadata fetching, which should significantly improve performance on large datasets. The code changes are well-structured, with logic nicely refactored into helper functions. I've identified a few minor issues, such as leftover debug statements and a TODO, which should be addressed for code hygiene. Overall, this is a great improvement.
alexeykudinkin
changed the title
ParquetDatasource
alexeykudinkin
force-pushed
the
ak/pq-ds-mtd-ftch-opt
branch
from
568d933 to
1c21d0a
Compare
…quetDatasource` Signed-off-by: Alexey Kudinkin <[email protected]>
Signed-off-by: Alexey Kudinkin <[email protected]>
Signed-off-by: Alexey Kudinkin <[email protected]>
Signed-off-by: Alexey Kudinkin <[email protected]>
Signed-off-by: Alexey Kudinkin <[email protected]>
Signed-off-by: Alexey Kudinkin <[email protected]>
alexeykudinkin
force-pushed
the
ak/pq-ds-mtd-ftch-opt
branch
from
1c21d0a to
74603f3
Compare
Signed-off-by: Alexey Kudinkin <[email protected]>
goutamvenkat-anyscale
approved these changes
Sep 2, 2025
Signed-off-by: Alexey Kudinkin <[email protected]>
8 tasks
alexeykudinkin
added a commit
that referenced
this pull request
Sep 6, 2025
<!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for #56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
sampan-s-nayak
pushed a commit
to sampan-s-nayak/ray
that referenced
this pull request
Sep 8, 2025
…-project#56105) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: sampan <[email protected]>
sampan-s-nayak
pushed a commit
to sampan-s-nayak/ray
that referenced
this pull request
Sep 8, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: sampan <[email protected]>
jugalshah291
pushed a commit
to jugalshah291/ray_fork
that referenced
this pull request
Sep 11, 2025
…-project#56105) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: jugalshah291 <[email protected]>
jugalshah291
pushed a commit
to jugalshah291/ray_fork
that referenced
this pull request
Sep 11, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: jugalshah291 <[email protected]>
wyhong3103
pushed a commit
to wyhong3103/ray
that referenced
this pull request
Sep 12, 2025
…-project#56105) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: yenhong.wong <[email protected]>
wyhong3103
pushed a commit
to wyhong3103/ray
that referenced
this pull request
Sep 12, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: yenhong.wong <[email protected]>
alexeykudinkin
added a commit
that referenced
this pull request
Sep 23, 2025
) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
alexeykudinkin
added a commit
that referenced
this pull request
Sep 23, 2025
<!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for #56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
ZacAttack
pushed a commit
to ZacAttack/ray
that referenced
this pull request
Sep 24, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: zac <[email protected]>
dstrodtman
pushed a commit
that referenced
this pull request
Oct 6, 2025
) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: Douglas Strodtman <[email protected]>
dstrodtman
pushed a commit
that referenced
this pull request
Oct 6, 2025
<!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for #56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]> Signed-off-by: Douglas Strodtman <[email protected]>
justinyeh1995
pushed a commit
to justinyeh1995/ray
that referenced
this pull request
Oct 20, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
snorkelopsstgtesting1-spec
pushed a commit
to snorkel-marlin-repos/ray-project_ray_pr_56299_264afdbd-dcea-472b-9966-b2fd4ffce1f5
that referenced
this pull request
Oct 22, 2025
<!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project/ray#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes #1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
snorkelopstesting2-coder
pushed a commit
to snorkel-marlin-repos/ray-project_ray_pr_56105_05e33cc6-0672-4d8a-826c-d64b6a3f24d1
that referenced
this pull request
Oct 22, 2025
Original PR #56105 by alexeykudinkin Original: ray-project/ray#56105
snorkelopstesting3-bot
added a commit
to snorkel-marlin-repos/ray-project_ray_pr_56105_05e33cc6-0672-4d8a-826c-d64b6a3f24d1
that referenced
this pull request
Oct 22, 2025
…ParquetDatasource` Merged from original PR #56105 Original: ray-project/ray#56105
bveeramani
added a commit
that referenced
this pull request
Oct 23, 2025
…se test This test became broken after the removal of Parquet metadata fetching tasks in #56105. The test relies on specific autoscaling behavior that no longer works as expected without metadata fetching. Signed-off-by: Balaji Veeramani <[email protected]>
bveeramani
added a commit
that referenced
this pull request
Oct 23, 2025
…ease test (#58048) ## Summary This PR removes the `image_classification_chaos_no_scale_back` release test and its associated setup script (`setup_cluster_compute_config_updater.py`). This test has become non-functional and is no longer providing useful signal. ## Background The `image_classification_chaos_no_scale_back` release test was designed to validate Ray Data's fault tolerance when many nodes abruptly get preempted at the same time. The test worked by: 1. Running on an autoscaling cluster with 1-10 nodes 2. Updating the compute config mid-test to downscale to 5 nodes 3. Asserting that there are dead nodes as a sanity check ## Why This Test Is Broken After the removal of Parquet metadata fetching in #56105 (September 2, 2025), the autoscaling behavior changed significantly: - **Before metadata removal**: The cluster would autoscale more aggressively because metadata fetching created additional tasks that triggered faster scale-up. The cluster would scale past 5 nodes, then downscale, leaving dead nodes that the test could detect. - **After metadata removal**: Without the metadata fetching tasks, the cluster doesn't scale up fast enough to get past 5 nodes before the downscale happens. This means there are no dead nodes to detect, causing the test to fail. ## Why We're Removing It 1. **Test is fundamentally broken**: The test's assumptions about autoscaling behavior are no longer valid after the metadata fetching removal 2. **Not actively monitored**: The test is marked as unstable and isn't closely watched ## Changes - Removed `image_classification_chaos_no_scale_back` test from `release/release_data_tests.yaml` - Deleted `release/nightly_tests/setup_cluster_compute_config_updater.py` (only used by this test) ## Related See #56105 Fixes #56528 Signed-off-by: Balaji Veeramani <[email protected]>
xinyuangui2
pushed a commit
to xinyuangui2/ray
that referenced
this pull request
Oct 27, 2025
…ease test (ray-project#58048) ## Summary This PR removes the `image_classification_chaos_no_scale_back` release test and its associated setup script (`setup_cluster_compute_config_updater.py`). This test has become non-functional and is no longer providing useful signal. ## Background The `image_classification_chaos_no_scale_back` release test was designed to validate Ray Data's fault tolerance when many nodes abruptly get preempted at the same time. The test worked by: 1. Running on an autoscaling cluster with 1-10 nodes 2. Updating the compute config mid-test to downscale to 5 nodes 3. Asserting that there are dead nodes as a sanity check ## Why This Test Is Broken After the removal of Parquet metadata fetching in ray-project#56105 (September 2, 2025), the autoscaling behavior changed significantly: - **Before metadata removal**: The cluster would autoscale more aggressively because metadata fetching created additional tasks that triggered faster scale-up. The cluster would scale past 5 nodes, then downscale, leaving dead nodes that the test could detect. - **After metadata removal**: Without the metadata fetching tasks, the cluster doesn't scale up fast enough to get past 5 nodes before the downscale happens. This means there are no dead nodes to detect, causing the test to fail. ## Why We're Removing It 1. **Test is fundamentally broken**: The test's assumptions about autoscaling behavior are no longer valid after the metadata fetching removal 2. **Not actively monitored**: The test is marked as unstable and isn't closely watched ## Changes - Removed `image_classification_chaos_no_scale_back` test from `release/release_data_tests.yaml` - Deleted `release/nightly_tests/setup_cluster_compute_config_updater.py` (only used by this test) ## Related See ray-project#56105 Fixes ray-project#56528 Signed-off-by: Balaji Veeramani <[email protected]> Signed-off-by: xgui <[email protected]>
landscapepainter
pushed a commit
to landscapepainter/ray
that referenced
this pull request
Nov 17, 2025
…-project#56105) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> Historically, `ParquetDatasource` have been fetching individual files parquet footer metadata to obtain granular metadata about the dataset. While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files). This change revisits this approach by - Removing metadata fetching as a step (and deprecating involved components) - Cleaning up and streamlining some of the other utilities further optimizing performance <!-- Please give a short summary of the change and the problem this solves. --> ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
landscapepainter
pushed a commit
to landscapepainter/ray
that referenced
this pull request
Nov 17, 2025
…ct#56268) <!-- Thank you for your contribution! Please review https://github.com/ray-project/ray/blob/master/CONTRIBUTING.rst before opening a pull request. --> <!-- Please add a reviewer to the assignee section when you create a PR. If you don't have the access to it, we will shortly find a reviewer and assign them to your PR. --> ## Why are these changes needed? This change is a follow-up for ray-project#56105. Now dataset size estimation is based on listed file sizes. However, encoding ratio was still based on the file size estimates derived from the uncompressed data size obtained from Parquet metadata. This change is addressing that by: - Rebasing encoding ratio to relate estimated in-memory size to the listed file size - Cleaning up unused abstractions (like `ParquetMetadataProvider`) ## Related issue number <!-- For example: "Closes ray-project#1234" --> ## Checks - [ ] I've signed off every commit(by using the -s flag, i.e., `git commit -s`) in this PR. - [ ] I've run `scripts/format.sh` to lint the changes in this PR. - [ ] I've included any doc changes needed for https://docs.ray.io/en/master/. - [ ] I've added any new APIs to the API Reference. For example, if I added a method in Tune, I've added it in `doc/source/tune/api/` under the corresponding `.rst` file. - [ ] I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/ - Testing Strategy - [ ] Unit tests - [ ] Release tests - [ ] This PR is not tested :( --------- Signed-off-by: Alexey Kudinkin <[email protected]>
landscapepainter
pushed a commit
to landscapepainter/ray
that referenced
this pull request
Nov 17, 2025
…ease test (ray-project#58048) ## Summary This PR removes the `image_classification_chaos_no_scale_back` release test and its associated setup script (`setup_cluster_compute_config_updater.py`). This test has become non-functional and is no longer providing useful signal. ## Background The `image_classification_chaos_no_scale_back` release test was designed to validate Ray Data's fault tolerance when many nodes abruptly get preempted at the same time. The test worked by: 1. Running on an autoscaling cluster with 1-10 nodes 2. Updating the compute config mid-test to downscale to 5 nodes 3. Asserting that there are dead nodes as a sanity check ## Why This Test Is Broken After the removal of Parquet metadata fetching in ray-project#56105 (September 2, 2025), the autoscaling behavior changed significantly: - **Before metadata removal**: The cluster would autoscale more aggressively because metadata fetching created additional tasks that triggered faster scale-up. The cluster would scale past 5 nodes, then downscale, leaving dead nodes that the test could detect. - **After metadata removal**: Without the metadata fetching tasks, the cluster doesn't scale up fast enough to get past 5 nodes before the downscale happens. This means there are no dead nodes to detect, causing the test to fail. ## Why We're Removing It 1. **Test is fundamentally broken**: The test's assumptions about autoscaling behavior are no longer valid after the metadata fetching removal 2. **Not actively monitored**: The test is marked as unstable and isn't closely watched ## Changes - Removed `image_classification_chaos_no_scale_back` test from `release/release_data_tests.yaml` - Deleted `release/nightly_tests/setup_cluster_compute_config_updater.py` (only used by this test) ## Related See ray-project#56105 Fixes ray-project#56528 Signed-off-by: Balaji Veeramani <[email protected]>
Aydin-ab
pushed a commit
to Aydin-ab/ray-aydin
that referenced
this pull request
Nov 19, 2025
…ease test (ray-project#58048) ## Summary This PR removes the `image_classification_chaos_no_scale_back` release test and its associated setup script (`setup_cluster_compute_config_updater.py`). This test has become non-functional and is no longer providing useful signal. ## Background The `image_classification_chaos_no_scale_back` release test was designed to validate Ray Data's fault tolerance when many nodes abruptly get preempted at the same time. The test worked by: 1. Running on an autoscaling cluster with 1-10 nodes 2. Updating the compute config mid-test to downscale to 5 nodes 3. Asserting that there are dead nodes as a sanity check ## Why This Test Is Broken After the removal of Parquet metadata fetching in ray-project#56105 (September 2, 2025), the autoscaling behavior changed significantly: - **Before metadata removal**: The cluster would autoscale more aggressively because metadata fetching created additional tasks that triggered faster scale-up. The cluster would scale past 5 nodes, then downscale, leaving dead nodes that the test could detect. - **After metadata removal**: Without the metadata fetching tasks, the cluster doesn't scale up fast enough to get past 5 nodes before the downscale happens. This means there are no dead nodes to detect, causing the test to fail. ## Why We're Removing It 1. **Test is fundamentally broken**: The test's assumptions about autoscaling behavior are no longer valid after the metadata fetching removal 2. **Not actively monitored**: The test is marked as unstable and isn't closely watched ## Changes - Removed `image_classification_chaos_no_scale_back` test from `release/release_data_tests.yaml` - Deleted `release/nightly_tests/setup_cluster_compute_config_updater.py` (only used by this test) ## Related See ray-project#56105 Fixes ray-project#56528 Signed-off-by: Balaji Veeramani <[email protected]> Signed-off-by: Aydin Abiar <[email protected]>
Future-Outlier
pushed a commit
to Future-Outlier/ray
that referenced
this pull request
Dec 7, 2025
…ease test (ray-project#58048) ## Summary This PR removes the `image_classification_chaos_no_scale_back` release test and its associated setup script (`setup_cluster_compute_config_updater.py`). This test has become non-functional and is no longer providing useful signal. ## Background The `image_classification_chaos_no_scale_back` release test was designed to validate Ray Data's fault tolerance when many nodes abruptly get preempted at the same time. The test worked by: 1. Running on an autoscaling cluster with 1-10 nodes 2. Updating the compute config mid-test to downscale to 5 nodes 3. Asserting that there are dead nodes as a sanity check ## Why This Test Is Broken After the removal of Parquet metadata fetching in ray-project#56105 (September 2, 2025), the autoscaling behavior changed significantly: - **Before metadata removal**: The cluster would autoscale more aggressively because metadata fetching created additional tasks that triggered faster scale-up. The cluster would scale past 5 nodes, then downscale, leaving dead nodes that the test could detect. - **After metadata removal**: Without the metadata fetching tasks, the cluster doesn't scale up fast enough to get past 5 nodes before the downscale happens. This means there are no dead nodes to detect, causing the test to fail. ## Why We're Removing It 1. **Test is fundamentally broken**: The test's assumptions about autoscaling behavior are no longer valid after the metadata fetching removal 2. **Not actively monitored**: The test is marked as unstable and isn't closely watched ## Changes - Removed `image_classification_chaos_no_scale_back` test from `release/release_data_tests.yaml` - Deleted `release/nightly_tests/setup_cluster_compute_config_updater.py` (only used by this test) ## Related See ray-project#56105 Fixes ray-project#56528 Signed-off-by: Balaji Veeramani <[email protected]> Signed-off-by: Future-Outlier <[email protected]>
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.
Historically,
ParquetDatasourcehave been fetching individual files parquet footer metadata to obtain granular metadata about the dataset.While laudable in principle, it's really inefficient in practice and manifests itself in extremely poor performance on very large datasets (10s to 100s Ks of files).
This change revisits this approach by
Related issue number
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.