This document will help to set up your Milvus development environment and to run tests. Please [file an issue](https://github.com/milvus-io/milvus/issues/new/choose) if there's a problem.
Our official Milvus versions are releases as Docker images. To build Milvus Docker on your own, please follow [these instructions](https://github.com/milvus-io/milvus/blob/master/build/README.md).
All Linux distributions are available for Milvus development. However a majority of our contributor worked with Ubuntu or CentOS systems, with a small portion of Mac (both x86_64 and Apple Silicon) contributors. If you would like Milvus to build and run on other distributions, you are more than welcome to file an issue and contribute!
- [Google Test](https://github.com/google/googletest.git) is automatically cloned from GitHub, which in some case could conflict with your local google test library.
The algorithm library of Milvus, Knowhere is written in c++. CMake is required in the Milvus compilation. If you don't have it, please follow the instructions in the [Installing CMake](https://cmake.org/install/).
Milvus is written in [Go](http://golang.org/). If you don't have a Go development environment, please follow the instructions in the [Go Getting Started guide](https://golang.org/doc/install).
Confirm that your `GOPATH` and `GOBIN` environment variables are correctly set as detailed in [How to Write Go Code](https://golang.org/doc/code.html) before proceeding.
Milvus depends on etcd, Pulsar and MinIO. Using Docker Compose to manage these is an easy way in local development. To install Docker and Docker Compose in your development environment, follow the instructions from the Docker website below:
Milvus uses `conan` to manage 3rd-party dependencies. `conan` will check the consistency of these dependencies every time you run `make`. This process can take a considerable amount of time, especially if the network is poor. If you make sure that the 3rd-party dependencies are consistent, you can use the following command to skip this step:
```shell
$ make SKIP_3RDPARTY=1
```
If this command succeeds, you will now have an executable at `bin/milvus` in your Milvus project directory.
If you want to run the `bin/milvus` executable on the host machine, you need to set `LD_LIBRARY_PATH` temporarily:
The PAGESIZE for the ky10 SP3 operating system is 65536, which is 64KB.
#### Modify the MILVUS_JEMALLOC_LG_PAGE setting
The `MILVUS_JEMALLOC_LG_PAGE` variable's primary function is to specify the size of large pages during the compilation of jemalloc. Jemalloc is a memory allocator designed to enhance the performance and efficiency of applications in a multi-threaded environment. By specifying the size of large pages, memory management and access can be optimized, thereby improving performance.
Large page support allows the operating system to manage and allocate memory in larger blocks, reducing the number of page table entries, thereby decreasing the time for page table lookups and improving the efficiency of memory access. This is particularly important when processing large amounts of data, as it can significantly reduce page faults and Translation Lookaside Buffer (TLB) misses, enhancing application performance.
On ARM64 architectures, different systems may support different page sizes, such as 4KB and 64KB. The `MILVUS_JEMALLOC_LG_PAGE` setting allows developers to customize the compilation of jemalloc for the target platform, ensuring it can efficiently operate on systems with varying page sizes. By specifying the `--with-lg-page` configuration option, jemalloc can utilize the optimal page size supported by the system when managing memory.
For example, if a system supports a 64KB page size, by setting `MILVUS_JEMALLOC_LG_PAGE` to the corresponding value (the power of 2, 64KB is 2 to the 16th power, so the value is 16), jemalloc can allocate and manage memory in 64KB units, which can improve the performance of applications running on that system.
Modify the make configuration file, located at: `./milvus/scripts/core_build.sh`, with the following changes:
Using `-DMILVUS_JEMALLOC_LG_PAGE=16` as a compilation option for jemalloc is because it specifies the size
of "large pages" as 2 to the 16th power bytes, which equals 65536 bytes or 64KB. This value is set to optimize memory management and improve performance, especially on systems that support or prefer using large pages to reduce the overhead of page table management.
Specifying `-DMILVUS_JEMALLOC_LG_PAGE=16` during the compilation of jemalloc informs jemalloc to assume the system's large page size is 64KB. This allows jemalloc to work more efficiently with the operating system's memory manager, using large pages to optimize performance. This is crucial for ensuring optimal performance on systems with different default page sizes, particularly in environments that might have different memory management needs due to varying hardware or system configurations.
### Build Image
```bash
cd ./milvus
cp build/docker/milvus/ubuntu20.04/Dockerfile .
```
Modify the Dockerfile as follows:
```dockerfile
# Copyright (C) 2019-2022 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance
# with the License. You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software distributed under the License
# is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express
# or implied. See the License for the specific language governing permissions and limitations under the License.
Pre-submission verification provides a battery of checks and tests to give your pull request the best chance of being accepted. Developers need to run as many verification tests as possible locally.
Milvus uses Python SDK to write test cases to verify the correctness of Milvus functions. Before running E2E tests, you need a running Milvus. There are two modes of operation to build Milvus — Milvus Standalone and Milvus Cluster. Milvus Standalone operates independently as a single instance. Milvus Cluster operates across multiple nodes. All milvus instances are clustered together to form a unified system to support larger volumes of data and higher traffic loads.
Both include three components:
1. Milvus: The core functional component.
2. Etcd: The metadata engine. Access and store metadata of Milvus’ internal components.
3. MinIO: The storage engine. Responsible for data persistence for Milvus.
Milvus Cluster includes further component — Pulsar, to be distributed through Pub/Sub mechanism.
A: We are aware that some tests can be flaky occasionally. If there's something you believe is abnormal (i.e. tests that fail every single time). You are more than welcome to [file an issue](https://github.com/milvus-io/milvus/issues/new/choose)!
