Repository navigation
[feat] Support cross-job actor discovery via explicit namespace - #115
Merged
Merged
Conversation
CLA Signature Guide@huniu20 , thanks for your pull request. The following commit(s) are not associated with a signed Contributor License Agreement (CLA).
To sign CLA, click here. To check if your email is configured correctly, refer to the FAQs. Once you've signed the CLA or updating your email, please comment |
ji-huazhong
self-requested a review
June 4, 2026 08:58
Collaborator
|
Thanks for your contribution. Kindly resolve failing CI tests first. @huniu20 |
huniu20
force-pushed
the
feat/cross-job-namespace
branch
from
June 4, 2026 09:09
5a1900d to
d02f41b
Compare
CLA Signature Guide@huniu20 , thanks for your pull request. The following commit(s) are not associated with a signed Contributor License Agreement (CLA).
To sign CLA, click here. To check if your email is configured correctly, refer to the FAQs. Once you've signed the CLA or updating your email, please comment |
Contributor
Author
|
/check-cla |
CLA Signature Passhuniu20, thanks for your pull request. All authors of the commits have signed the CLA. 👍 |
huniu20
force-pushed
the
feat/cross-job-namespace
branch
from
June 4, 2026 09:51
d02f41b to
9af6f55
Compare
CLA Signature Passhuniu20, thanks for your pull request. All authors of the commits have signed the CLA. 👍 |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR makes TransferQueue’s Ray named actors discoverable across multiple Ray Jobs sharing the same cluster by consistently using a fixed Ray namespace ("transfer_queue") when creating and retrieving the controller (and the Ray storage actor).
Changes:
- Add
namespace="transfer_queue"toray.get_actor("TransferQueueController", ...)call sites (library + tests + examples). - Create
TransferQueueController(andRayObjectRefStorage) explicitly in the"transfer_queue"namespace via.options(namespace=...). - Update E2E tests and tutorial/demo code to retrieve the controller from the fixed namespace.
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 |
|---|---|
| tutorial/06_streaming_dataloader.py | Fetches the controller from the fixed Ray namespace in the tutorial worker path. |
| transfer_queue/storage/clients/ray_storage_client.py | Fetches/creates RayObjectRefStorage in the fixed Ray namespace to enable cross-job access. |
| transfer_queue/interface.py | Fetches/creates TransferQueueController in the fixed Ray namespace for cross-job discovery. |
| tests/e2e/test_kv_interface_e2e.py | Updates test fixture to retrieve controller from the fixed namespace. |
| recipe/simple_use_case/relax_demo.py | Updates demo worker to retrieve controller from the fixed namespace. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| try: | ||
| if _TQ_CONTROLLER is None: | ||
| _TQ_CONTROLLER = ray.get_actor("TransferQueueController") | ||
| _TQ_CONTROLLER = ray.get_actor("TransferQueueController", namespace="transfer_queue") |
| # initialize actor | ||
| try: | ||
| self.storage_actor = ray.get_actor("RayObjectRefStorage") | ||
| self.storage_actor = ray.get_actor("RayObjectRefStorage", namespace="transfer_queue") |
When multiple Ray Jobs share the same Ray cluster, Named Actors are isolated by namespace. Without an explicit namespace, a TQ Controller created by one job is invisible to workers in another job. This commit adds namespace="transfer_queue" to both: - ray.get_actor() in _init_from_existing() - TransferQueueController.options() in init() This ensures that the TQ Controller is always registered and discovered in the fixed "transfer_queue" namespace, enabling cross-job TQ sharing (e.g., a teacher server job creates TQ, and a trainer job connects to it). This change is backward-compatible: single-job usage is unaffected since the namespace is consistent between creation and discovery. Signed-off-by: huniu20 <huniumail@gmail.com>
huniu20
force-pushed
the
feat/cross-job-namespace
branch
from
June 4, 2026 10:06
9af6f55 to
b20656a
Compare
CLA Signature Passhuniu20, thanks for your pull request. All authors of the commits have signed the CLA. 👍 |
0oshowero0
approved these changes
Jun 5, 2026
0oshowero0
pushed a commit
that referenced
this pull request
Oct 30, 2026
…TQ_RAY_NAMESPACE (#206) ## Problem The controller, `RayObjectRefStorage` and the `TransferQueueLockManager_*` actors are named actors in the fixed Ray namespace `transfer_queue`. Two jobs on one Ray cluster (for example two `ray job submit` runs of a trainer) therefore attach to the same controller: - the second job's `tq.init()` finds the first job's controller and reads its partitions; - the first job's `tq.close()` kills that controller while the second job still uses it. #115 made the namespace fixed on purpose, so a trainer job can attach to TransferQueue created by another job. That behaviour is kept; this PR only makes the namespace selectable. ## Change `ray_namespace()` in `utils/common.py` returns `TQ_RAY_NAMESPACE`, or `transfer_queue` when it is unset or empty. It replaces the literal at all six sites (controller create and lookup, Ray storage actor create and lookup, lock manager create and lookup). A job that must stay isolated sets the variable to a per-job value in every process (for example through the job-level `runtime_env` `env_vars`). The `ray_namespace()` docstring documents it. The default is unchanged, so nothing changes for existing users. ## Evidence Two drivers on one local Ray head (`ray start --head`), the first one puts `("p", "k")` and stays alive, the second one calls `tq.kv_list("p")`: | Setup | Second driver sees | | --- | --- | | no `TQ_RAY_NAMESPACE` | `{'p': {'k': {}}}` (the first job's data) | | `TQ_RAY_NAMESPACE=job-a` / `job-b` | `{}` | ## Tests - `pytest tests/test_kv_lock.py tests/test_storage_client_factory.py tests/e2e/test_kv_global_lock_e2e.py`: 30 passed, 2 skipped, Ray 2.57.0, torch 2.13, Python 3.12. - `ruff check` and `ruff format --check` clean on the changed files. --------- Signed-off-by: nataliekung <4054699+nataliekung@users.noreply.github.com> Co-authored-by: nataliekung <4054699+nataliekung@users.noreply.github.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
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.
When multiple Ray Jobs share the same Ray cluster, Named Actors are isolated by namespace. Without an explicit namespace, a TQ Controller created by one job is invisible to workers in another job.
This commit adds namespace="transfer_queue" to both:
This ensures that the TQ Controller is always registered and discovered in the fixed "transfer_queue" namespace, enabling cross-job TQ sharing (e.g., a teacher server job creates TQ, and a trainer job connects to it).
This change is backward-compatible: single-job usage is unaffected since the namespace is consistent between creation and discovery.