Add client support for replication scale plans (GET/POST /v1/replication/scale)

Open
#2,259 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
48/100
Issue type
Feature
Clarity
Clearly specified
Activity status
Active
Tech stack
python
Domain
api, backend

Research direction

Read the existing client.cluster.replications methods (get, query and cancel), related models in weaviate.cluster.models, and the client’s version checks and stub generation workflow. Implement sync and async methods for retrieving and applying scale plans, with typed models, validation, and server version handling. Add unit, mock, and integration coverage; done means the scale plan can be retrieved and applied against a supported server, with the requested error and validation cases covered.

Written by the indexing model from the issue text.

Description

Summary

Weaviate has offered GET and POST /v1/replication/scale since v1.35. They compute and apply a plan to change a collection's replication factor. The Python client has no methods for them, so changing a collection's replication factor from Python means making raw REST calls and building the plan payload by hand.

Proposal

Add two methods to the existing client.cluster.replications namespace, next to get, query and cancel, for both the sync and async clients:

plan = client.cluster.replications.get_scale_plan(collection="Article", replication_factor=3)
# inspect plan.shard_scale_actions -> {shard: ShardScaleActions(remove_nodes=[...], add_nodes={target: source | None})}
result = client.cluster.replications.apply_scale_plan(plan=plan)
# result.operation_ids -> the started shard copy operations, trackable with replications.get(uuid=...)
  • New typed models in weaviate.cluster.models: ReplicationScalePlan, ShardScaleActions and ReplicationScaleApplyResult. They are exported from weaviate.outputs.replication.
  • A version check raises WeaviateUnsupportedFeatureError on servers older than 1.35, the same way tokenization and export do.
  • replication_factor is validated on the client (it must be an integer of at least 1), and the collection name's first letter is capitalised, as elsewhere in the client.
  • Additive only: no changes to existing signatures, overloads or dependencies. The stubs are regenerated with tools/stubs_regen.sh.
Validation done locally

I have a working implementation, tested against a local 3-node 1.40.5 cluster with REPLICA_MOVEMENT_ENABLED=true:

  • Scaling a 2-shard collection from 1 to 3 replicas started 4 copy operations, all of which reached READY, and all objects stayed readable.
  • Scaling back down produced a removal-only plan, applied with no copy operations.
  • Added unit and mock tests (sync and async, plan round-trip, invalid factor, old-server error) and an integration test in integration/test_replicate.py.
  • ruff, pyright 1.1.399 and the flake8 pre-commit hooks pass.

Would you accept a PR for this? I'm happy to adjust the method names or where they live, for example on client.cluster next to replicate() instead.

Dominant language
Python
Stars
230
Forks
173
Avg merge
4d 16h
Merged PRs (30d)
23

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from weaviate/weaviate-python-client

All issues in weaviate/weaviate-python-client

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.