Comments (2)
We have a use case for this, where certain classes that facilitate running work on a cluster internally use actors for state management (as an implementation detail). It is possible that several of these classes are instantiated as part of the same job, and in this case, we want them to be using the same actor. We have been using named actors for this (not detached), and the classes have been constructing them passing a common name and with the get_if_exists=True
flag set so they will be shared.
The change here makes the existing option less viable, as any classes that get an already-existing actor are only holding a weak reference to it, so if the original class that created it goes out of scope, the actor gets killed. Having an actor lifetime tied to the job/driver lifetime would be safer way to implement the original pattern.
from ray.
The use case sounds a bit narrow when applied in user-facing scenarios, can someone please provide some more use cases to when a job lifetime actor is very valuable?
from ray.
Related Issues (20)
- [data] ray.data.read_parquet not reading in one file per block HOT 3
- [Core, Observabilty] Create metric for running ray jobs HOT 1
- [RayTune+RayTrain]The PlacementGroupFactory in Ray Tune fails when using Ray XGBoost trainer
- [RayTune+RayTrain]Ray Tune results unpredictable when concurrent_trials > 1 but resources support only 1 trial
- Release test dataset_shuffle_sort_1tb.aws failed HOT 1
- Release test aws_cluster_launcher_full failed HOT 4
- [<Ray component: data>] Ray data can not handle big dataset.
- deploy ray on alibaba cloud vm failed HOT 3
- [Ray dashboard] Actors tab does not list actors under certain conditions
- pyarrow.lib.ArrowInvalid: offset overflow while concatenating arrays [<Ray component: Data>] HOT 1
- [Serve] @serve.batch decorator order issue HOT 2
- CI test linux://rllib:learning_tests_pendulum_ppo_gpu is consistently_failing HOT 6
- Release test dask_on_ray_100gb_sort.aws failed HOT 1
- [Core] Enormous file handler and memory leaks on windows HOT 2
- Global Per-Epoch Shuffling with TorchTrainer
- Release test read_images_train_4_gpu.aws failed HOT 3
- [Tensorflow] Trainer example does not run
- CI test linux://rllib:learning_tests_multi_agent_cartpole_ppo_multi_cpu is flaky HOT 8
- [Core] Runtime envs race condition when uploading packages
- CI test linux://rllib:learning_tests_stateless_cartpole_impala is consistently_failing HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from ray.