[core] Use core worker client pool in GCS#53654
Merged
jjyao merged 4 commits intoray-project:masterfrom Jun 21, 2025
Merged
Conversation
Signed-off-by: dayshah <dhyey2019@gmail.com>
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR refactors the GCS components to replace the per‐call core worker client factory with a shared core worker client pool, enabling the use of retryable gRPC clients for interactions between GCS and workers. Key changes include:
- Replacing all client_factory usages with a shared worker_client_pool across tests and production code.
- Updating GCS server components (job manager, actor scheduler, actor manager, etc.) to use the new worker_client_pool API.
- Adjusting test cases in various modules to work with the shared pool.
Reviewed Changes
Copilot reviewed 14 out of 14 changed files in this pull request and generated no comments.
Show a summary per file
| File | Description |
|---|---|
| src/ray/gcs/gcs_server/test/gcs_job_manager_test.cc | Updated to use worker_client_pool_ instead of client_factory. |
| src/ray/gcs/gcs_server/test/gcs_actor_scheduler_test.cc | Modified to instantiate and use worker_client_pool_ accordingly. |
| src/ray/gcs/gcs_server/test/gcs_actor_scheduler_mock_test.cc | Refactored to inject worker_client_pool_ in place of direct client creation. |
| src/ray/gcs/gcs_server/test/gcs_actor_manager_test.cc | Updated to use worker_client_pool_ for actor manager tests. |
| src/ray/gcs/gcs_server/test/export_api/gcs_actor_manager_export_event_test.cc | Modified to use worker_client_pool_ in export event tests. |
| src/ray/gcs/gcs_server/gcs_worker_manager.cc | Replaced uses of a temporary client_factory with appropriate worker_client_pool_ calls. |
| src/ray/gcs/gcs_server/gcs_server.h | Added include and member for the core worker client pool. |
| src/ray/gcs/gcs_server/gcs_server.cc | Refactored worker pool initialization and its usage in job and actor managers. |
| src/ray/gcs/gcs_server/gcs_job_manager.h & .cc | Updated constructor and calls to use the worker_client_pool. |
| src/ray/gcs/gcs_server/gcs_actor_scheduler.h & .cc | Replaced core_worker_client factory usage with the shared worker_client_pool_. |
| src/ray/gcs/gcs_server/gcs_actor_manager.h & .cc | Revised parameters and calls to use worker_client_pool_ for actor communication. |
Comments suppressed due to low confidence (1)
src/ray/gcs/gcs_server/gcs_job_manager.cc:370
- Previously a disconnect call was made when a job was marked dead. Verify that omitting worker_client_pool_.Disconnect(worker_id) here is intentional and that stale connections are properly handled elsewhere.
reply->mutable_job_info_list(jj)->set_is_running_tasks(false);
edoakes
approved these changes
Jun 9, 2025
Collaborator
israbbani
reviewed
Jun 9, 2025
Signed-off-by: dayshah <dhyey2019@gmail.com>
Contributor
|
How do we GC clients inside the pool? |
dayshah
commented
Jun 10, 2025
| [this](const std::shared_ptr<rpc::WorkerTableData> &worker_failure_data) { | ||
| auto &worker_address = worker_failure_data->worker_address(); | ||
| auto worker_id = WorkerID::FromBinary(worker_address.worker_id()); | ||
| worker_client_pool_.Disconnect(worker_id); |
Contributor
There was a problem hiding this comment.
We also need it when the node dies, since if node dies, we won't call this listener for each individual worker on the node.
Contributor
Author
There was a problem hiding this comment.
oh I expected the worker death to be propagated there in case of node death too.
Anyways updated to disconnect on node death too
Signed-off-by: dayshah <dhyey2019@gmail.com>
Signed-off-by: dayshah <dhyey2019@gmail.com>
minerharry
pushed a commit
to minerharry/ray
that referenced
this pull request
Jun 27, 2025
Signed-off-by: dayshah <dhyey2019@gmail.com>
elliot-barn
pushed a commit
that referenced
this pull request
Jul 2, 2025
Signed-off-by: dayshah <dhyey2019@gmail.com> Signed-off-by: elliot-barn <elliot.barnwell@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
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.
Why are these changes needed?
Replacing all uses of a core worker client factory / unshared client pool with a shared core worker client pool for the whole gcs server. This is needed to use the retryable grpc client in the gcs. Will be implementing retries for rpc's that exist and new gcs -> worker rpc's will need retries, ex. #51653
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.