Serve relies on being able to do quiet application-level retries, and this info-level logging is resulting in log spam hitting users. This PR demotes this log statement to debug-level to prevent this log spam.
Co-authored-by: simon-mo <simon.mo@hey.com>
This PR is to improve in-memory data size estimation of image folder data source. Before this PR, we use on-disk file size as estimation of in-memory data size of image folder data source. This can be inaccurate due to image compression and in-memory image resizing.
Given `size` and `mode` is set to be optional in https://github.com/ray-project/ray/pull/27295, so change this PR to tackle the simple case when `size` and `mode` are both provided.
* `size` and `mode` is provided: just calculate the in-memory size based on the dimensions, not need to read any image (this PR)
* `size` or `mode` is not provided: need sampling to determine the in-memory size (will do in another followup PR).
Here is example of estiamted size for our test image folder
```
>>> import ray
>>> from ray.data.datasource.image_folder_datasource import ImageFolderDatasource
>>> root = "example://image-folders/different-sizes"
>>> ds = ray.data.read_datasource(ImageFolderDatasource(), root=root, size=(64, 64), mode="RGB")
>>> ds.size_bytes()
40310
>>> ds.fully_executed().size_bytes()
37428
```
Without this PR:
```
>>> ds.size_bytes()
18978
```
Adds a page describing a development workflow for Serve applications.
Co-authored-by: Edward Oakes <ed.nmi.oakes@gmail.com>
Co-authored-by: shrekris-anyscale <92341594+shrekris-anyscale@users.noreply.github.com>
Co-authored-by: Stephanie Wang <swang@cs.berkeley.edu>
The "Monitoring Ray Serve" page explains how to inspect your Ray Serve applications. This change updates the page to remove outdated metrics that Serve no longer exposes and to upgrade code samples to use 2.0 APIs. It also improves the content's readability and organization.
Link to updated "Monitoring Ray Serve" page: https://ray--27777.org.readthedocs.build/en/27777/serve/monitoring.html
Refactor Datasets API docs for easier navigation: [Ray Datasets API](https://ray--27592.org.readthedocs.build/en/27592/data/api/api.html)
### Changes
1. Create a new Datasets API base page.
2. Split existing APIs into separate pages.
3. Split `Dataset` and `DatasetPipeline` methods into separate sections.
1. Used `autosummary` to generate overview tables at the top of each of these pages. Open to other suggestions e.g. moving the summary to the top of each section instead.
2. **Note:** Every time we add a new method we need to explicitly add it here as well.
4. Add Input/Output APIs.
1. I chose to split these primarily by data format rather than type, since it's easier to navigate, and the existing [Creating Datasets](https://docs.ray.io/en/master/data/creating-datasets.html) User Guide already does the latter.
6. Add `Block` and `DataBatch` (should we add these aliases?)
7. Remove existing `package-ref`.
The actor handle held at Ray client will become dangling if the Ray cluster is shutdown, and in such case if the user tries to get the actor again it will result in crash. This happened in a real user and blocked them from making progress.
This change makes the stats actor detached, and instead of keeping a handle, we access it via its name. This way we can make sure re-create this actor if the cluster gets restarted.
Co-authored-by: Ubuntu <ubuntu@ip-172-31-32-136.us-west-2.compute.internal>
Page structure changes:
Deploying a Ray Cluster on Kubernetes
Getting Started -> links to jobs
Deploying a Ray Cluster on VMs
Getting started -> links to jobs
User Guides
Autoscaling (moved more content here in favor of the Getting started page)
Running Applications on Ray Clusters
Ray Jobs
Quickstart Using the Ray Jobs CLI
Python SDK
REST API
Ray Job Submission API Reference
Ray Client
Content changes:
modified "Deploying a Ray Cluster ..." quickstart pages to briefly summarize ad-hoc command execution, then link to jobs
modified Ray Jobs example to be more incremental - start with a simple example, then show long-running script, then show example with a runtime env, instead of all of them at once
center Ray Jobs quickstart around using the CLI. Made some minor changes to the Python SDK page to match it
remove "Ray Jobs Architecture"
moved "Autoscaling" content away from Kubernetes "Getting started" page into its own user guide. I think it's too complicated for "Getting Started". No content cuts.
Cut "Viewing the dashboard" and "Ray Client" from Kubernetes "Getting started" page.
Signed-off-by: Stephanie Wang <swang@cs.berkeley.edu>
A new feature was recently added, where Serve replicas are not restarted if only `num_replicas`, `autoscaling_config`, and/or `user_config` is updated in the config file that's redeployed. Updating docs to talk about this feature.
Co-authored-by: shrekris-anyscale <92341594+shrekris-anyscale@users.noreply.github.com>
When the node id of the controller died, GSC will try to reschedule the controller to the same node. But GCS will only mark the node as failure after 120s when GCS restarts (or 30s if only raylet died).
This PR fixed it by unpin it to the head node. So as long as GCS is alive, it'll reschedule it immediately. But we can't turn it on by default, so we introduce an internal flag for this.
