Files
milvus/tests/python_client
aea5d701bd enhance: bind index meta to add function field DDL (#51105)
issue: #51104
companion SDK PR (merge first): milvus-io/pymilvus#3670

## Summary

`add_function_field` previously accepted a vector output field without
any index meta; bump-schema-version compaction then stalled forever (no
segment index task can be created for a field with no index defined, and
`GetQueryVChanPositions` never hands off unindexed compacted-to segments
— the fail-closed chain is silent). This PR makes the index meta a
mandatory, atomically-bound part of the DDL:

- **Mandatory at every layer**: SDK requires `index_params`; proxy and
rootcoord reject vector function-output fields without an explicit
`index_type` strictly before the broadcast (no silent AUTOINDEX).
- **create_index-aligned materialization**: proxy normalizes the params
with the same logic as `create_index` (name rules via
`validateIndexName`, checker existence, `CheckTrain` incl. dimension
filling, function-type defaults such as `bm25_k1/bm25_b/bm25_avgdl`) and
writes them back into the request; rootcoord prepare allocates index
id/name, rejects name conflicts and unknown index types, and serializes
a complete `indexpb.FieldIndex` into the new
`AlterCollectionMessageUpdates.bound_field_indexes` WAL field.
- **Atomic apply**: the alter-collection ack callback commits the schema
and then applies the bound index by dispatching a synthetic
`CreateIndexMessage` to datacoord's existing `createIndexV2AckCallback`
(same pattern as `cascadeDropFieldIndexesInline`). The callback is
replayed until success across crashes and is fully idempotent (all ids
come from the message body), so `DDL success ⇒ index meta exists` holds
with no partial terminal state.
- **Reuse promotion**: `ValidateIndexParams`/`CheckDuplidateKey` moved
from datacoord to `internal/util/indexparamcheck`; shared
`ExpandIndexParams` / `FillFunctionOutputIndexParams` /
`PrepareFunctionOutputIndexParams` helpers now back both the
create_index path and this DDL.
- **Dead fan-out removed**: the ordinary `CreateIndex` broadcast is
reverted to control-channel-only; the vchannel fan-out had no consumer
and wrote one dead WAL entry per vchannel per index creation.
- **No milvus-proto change**: the request already carried
`FieldInfo.index_name`/`extra_params` (previously unused); this feature
activates them.
- e2e suites updated to the new semantics:
`test_add_function_field_feature.py` rewritten (explicit `index_params`
everywhere, post-DDL `create_index` blocks removed since the bound index
conflicts with a second distinct index per existing semantics, new
negative cases for the mandatory checks); the external-table negative
case updated as well.

## Verification

- New/updated unit tests: rootcoord DDL callbacks
(mandatory/AUTOINDEX/unknown-index-type rejections, bound-index
materialization + ack-callback application via a recording fake), proxy
task validation (invalid index name, normalization write-back), shared
helper tests.
- End-to-end on a live cluster (StorageV3 + bump compaction enabled):
request without params rejected; index meta visible with
create_index-aligned params immediately after the DDL returns; bump
compaction rewrote 2000-row sealed segments and the bound index built on
the results (the exact chain that previously stalled); BM25 search over
pre-DDL rows succeeds once the compacted segments are loaded.

## Limitations (follow-ups)

- Live propagation of the new index meta to already-loaded QueryNode
delegators is intentionally out of scope; a loaded collection observes
the bound index through the segment load path (e.g. reload /
compacted-segment load). Tracked as a separate follow-up PR.
- The python e2e suites require the companion pymilvus change:
milvus-io/pymilvus#3670 is merged and
`tests/python_client/requirements.txt` is bumped to
`pymilvus==3.1.0rc61` in this PR (all 22 rewritten cases + the
external-table case verified locally against the released rc61).
- During a rolling-upgrade window an old coordinator replaying the new
message ignores `bound_field_indexes` (proto3 unknown field), i.e.
pre-existing behavior; avoid the new argument in mixed-version clusters.
- `AddCollectionFunction` on an existing field and plain
add-vector-field keep their current behavior (the WAL carrier is generic
for future extension).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Signed-off-by: MrPresent-Han <chun.han@gmail.com>
Co-authored-by: MrPresent-Han <chun.han@gmail.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
2026-07-09 17:38:41 +08:00
..

Guidelines for Test Framework

This document guides you through the Pytest-based PyMilvus test framework.

You can find the test code on GitHub.

Quick Start

Deploy Milvus

To accommodate the variety of requirements, Milvus offers as many as four deployment methods. PyMilvus supports Milvus deployed with any of the methods below:

  1. Build from source code

  2. Install with Docker Compose

  3. Install on Kunernetes

  4. Install with KinD

For test purposes, we recommend installing Milvus with KinD. KinD supports the ClickOnce deployment of Milvus and its test client. KinD deployment is tailored for scenarios with small data scale, such as development/debugging test cases and functional verification.

  • Prerequisites

  • Install KinD with script

    1. Enter the local directory of the code */milvus/tests/scripts/
    2. Build the KinD environment, and execute CI Regression test cases automatically:
    $ ./e2e-k8s.sh 
    
    • By default, KinD environment will be automatically cleaned up after the execution of the test case. If you need to keep the KinD environment:
    $ ./e2e-k8s.sh --skip-cleanup
    
    • Skip the automatic test case execution and keep the KinD environment:
    $ ./e2e-k8s.sh --skip-cleanup --skip-test --manual
    

Note: You need to log in to the containers of the test client to proceed manual execution and debugging of the test case.

  • See more script parameters:
$ ./e2e-k8s.sh --help
  • Export cluster logs:
$ kind export logs .

PyMilvus Test Environment Deployment and Case Execution

We recommend using Python 3.12, consistent with the Python client CI runtime.

Note: Procedures listed below will be completed automatically if you deployed Milvus using KinD.

  1. Install the Python package prerequisite for the test, enter */milvus/tests/python_client/, and execute:

    $ pip install -r requirements.txt
    
  2. The default test log path is /tmp/ci_logs/ under the config directory. You can add environment variables to change the path before booting up test cases:

    $ export CI_LOG_PATH=/tmp/ci_logs/test/
    
Log Level Log File
debug ci_test_log.debug
info ci_test_log.log
error ci_test_log.err
  1. You can configure default parameters in pytest.ini under the root path. For instance:

    addopts = --host *.*.*.*  --html=/tmp/ci_logs/report.html
    

where host should be set as the IP address of the Milvus service, and *.html is the report generated for the test.

  1. Enter testcases directory, run following command, which is consistent with the command under the pytest framework, to execute the test case:

    $ python3 -W ignore -m pytest <test_file_name>
    

An Introduction to Test Modules

Module Overview

Module Overview

SDK Test Flow Chart

Working directories and files

  • base: stores the encapsulated PyMilvus module files, and setup & teardown functions for pytest framework.
  • check: stores the check module files for returned results from interface.
  • common: stores the files of common methods and parameters for test cases.
  • config: stores the basic configuration file.
  • testcases: stores test case scripts.
  • utils: stores utility programs, such as utility log and environment detection methods.
  • requirements.txt: specifies the python package required for executing test cases
  • conftest.py: you can compile fixture functions or local plugins in this file. These functions and plugins implement within the current folder and its subfolder.
  • pytest.ini: the main configuration file for pytest.

Critical design ideas

  • base/*_wrapper.py encapsulates the tested interface, uniformly processes requests from the interface, abstracts the returned results, and passes the results to check/func_check.py module for checking.
  • check/func_check.py encompasses result checking methods for each interface for invocation from test cases.
  • base/client_base.py uses pytest framework to process setup/teardown functions correspondingly.
  • Test case files in testcases folder should be compiled inheriting the TestcaseBase module from base/client_base.py. Compile the common methods and parameters used by test cases into the Common module for invocation.
  • Add global configurations under config directory, such as log path, etc.
  • Add global implementation methods under utils directory, such as utility log module.

Adding codes

This section specifies references while adding new test cases or framework tools.

Notice and best practices

  1. Coding style
  • Test files: each SDK category corresponds to a test file. So do load and search methods.

  • Test categories: test files fall into two categories

    • TestObjectParams:
      • Indicates the parameter test of corresponding interface. For instance, TestPartitionParams represents the parameter test for Partition interface.
      • Tests the target category/method under different parameter inputs. The parameter test will cover default, empty, none, datatype, maxsize, etc.
    • TestObjectOperations:
      • Indicates the function/operation test of corresponding interface. For instance, TestPartitionOperations represents the function/operation test for Partition interface.
      • Tests the target category/method with legit parameter inputs and interaction with other interfaces.
  • Testcase naming

    • TestObjectParams:

      • Name after the parameter input of the test case. For instance, test_partition_empty_name() represents test on performance with the empty string as the name parameter input.
    • TestObjectOperations

      • Name after the operation procedure of the test case. For instance, test_partition_drop_partition_twice() represents the test on the performance when dropping partitions twice consecutively.
      • Name after assertions. For instance, test_partition_maximum_partitions() represents test on the maximum number of partitions that can be created.
  1. Notice
  • Do not initialize PyMilvus objects in the test case files.
  • Generally, do not add log IDs to test case files.
  • Directly call the encapsulated methods or attributes in test cases, as shown below:

To create multiple partitions objects, call self.init_partition_wrap(), which returns the newly created partition objects. Call self.partition_wrap instead when you do not need multiple objects.

# create partition  -Call the default initialization method
partition_w = self.init_partition_wrap()
assert partition_w.is_empty
# create partition    -Directly call the encapsulated object
self.partition_wrap.init_partition(collection=collection_name, name=partition_name)
assert self.partition_wrap.is_empty
  • To test on the error or exception returned from interfaces:

    • Call check_task=CheckTasks.err_res.
    • Input the expected error ID and message.
    # create partition with collection is None
    self.partition_wrap.init_partition(collection=None, name=partition_name, check_task=CheckTasks.err_res, check_items={ct.err_code: 1, ct.err_msg: "'NoneType' object has no attribute"})
    
  • To test on the normal value returned from interfaces:

    • Call check_task=CheckTasks.check_partition_property. You can build new test methods in CheckTasks for invocation in test cases.
    • Input the expected result for test methods.
    # create partition
    partition_w = self.init_partition_wrap(collection_w, partition_name, check_task=CheckTasks.check_partition_property, check_items={"name": partition_name, "description": description, "is_empty": True, "num_entities": 0})
    
  1. Adding test cases
  • Find the encapsulated tested interface with the same name in the *_wrapper.py files under base directory. Each interface returns a list with two values, among which one is interface returned results of PyMilvus, and the other is the assertion of normal/abnormal results, i.e. True/False. The returned judgment can be used in the extra result checking of test cases.

  • Add the test cases in the corresponding test file of the tested interface in testcases folder. You can refer to all test files under this directory to create your own test cases as shown below:

    @pytest.mark.tags(CaseLabel.L1)
    @pytest.mark.parametrize("partition_name", [cf.gen_unique_str(prefix)])
    def test_partition_dropped_collection(self, partition_name):
        """
        target: verify create partition against a dropped collection
        method: 1. create collection1
                2. drop collection1
                3. create partition in collection1
        expected: raise exception
        """
        # create collection
        collection_w = self.init_collection_wrap()
        # drop collection
        collection_w.drop()
        # create partition failed
        self.partition_wrap.init_partition(collection_w.collection, partition_name, check_task=CheckTasks.err_res, check_items={ct.err_code: 4, ct.err_msg: "collection not found"})
    
  • Tips

    • Case comments encompass three parts: object, test method, and expected result. You should specify each part.
    • Initialize the tested category in the setup method of the Base category in the base/client_base.py file, as shown below:
    self.connection_wrap = ApiConnectionsWrapper()
    self.utility_wrap = ApiUtilityWrapper()
    self.collection_wrap = ApiCollectionWrapper()
    self.partition_wrap = ApiPartitionWrapper()
    self.index_wrap = ApiIndexWrapper()
    self.collection_schema_wrap = ApiCollectionSchemaWrapper()
    self.field_schema_wrap = ApiFieldSchemaWrapper()
    
    • Pass the parameters with corresponding encapsulated methods when calling the interface you need to test on. As shown below, align all parameters with those in PyMilvus interfaces except for check_task and check_items.
    def init_partition(self, collection, name, description="", check_task=None, check_items=None, **kwargs)
    
    • check_task is used to select the corresponding interface test method in the ResponseChecker check category in the check/func_check.py file. You can choose methods under the CheckTasks category in the common/common_type.py file.
    • The specific content of check_items passed to the test method is determined by the implemented test method check_task.
    • The tested interface can return normal results when CheckTasks and check_items are not passed.
  1. Adding framework functions
  • Add global methods or tools under utils directory.

  • Add corresponding configurations under config directory.