[RELEASE] v22.06 raft #708
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
Fix merge conflicts [skip gpuci]
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
This PR updates the `ucx-py` version for `branch-22.06`. Authors: - AJ Schmidt (https://github.com/ajschmidt8) Approvers: - Dante Gama Dessavre (https://github.com/dantegd) - Mark Sadang (https://github.com/msadang) - Corey J. Nolet (https://github.com/cjnolet) URL: #596
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
This PR allows building libcugraphops static libs via CMake option [`-DBUILD_SHARED_LIBS=ON|OFF`](https://cmake.org/cmake/help/latest/variable/BUILD_SHARED_LIBS.html). ~I was seeing a linker error due to not having `-fPIC` enabled, but I wouldn't have expected this to affect static libs. I'll investigate some more and turn it off if possible, but for now `-fPIC` is enabled.~ Authors: - Paul Taylor (https://github.com/trxcllnt) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #602
…604) This PR fixes issue - #568 and rapidsai/cuml#4624 -- fix issue in fusedL2knn which happens when rows are multiple of 256. -- make index value to be size_t to avoid int overflow though this doesn't hamper these issues but it may for higher input sizes. -- also add some additional test cases in fusedL2knn test. Authors: - Mahesh Doijade (https://github.com/mdoijade) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #604
[gpuCI] Forward-merge branch-22.04 to branch-22.06 [skip gpuci]
Authors: - Corey J. Nolet (https://github.com/cjnolet) Approvers: - Divye Gala (https://github.com/divyegala) URL: #613
[REVIEW] Unpin `dask` & `distributed` for development
CMake 3.23.1 contains the bug fixes that we need to make use of CMake 3.23, so now we can update the pinnings to just avoid 3.23.0. Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Jordan Jacobelli (https://github.com/Ethyling) - Robert Maynard (https://github.com/robertmaynard) URL: #617
In addition to the cuco dependency, the following changes are included: 1. ability to turn off thrust and mdspan dependencies (rmm is still required) 2. compiling libraries now defaults to the same setting of `BUILD_TESTS` (tests are still enabled) 3. cuco dependency is disabled by default (unless distance component is enabled) 4. the headers which are safe to expose in public APIs are moved over to `core/` directory. Authors: - Corey J. Nolet (https://github.com/cjnolet) Approvers: - Dante Gama Dessavre (https://github.com/dantegd) - Robert Maynard (https://github.com/robertmaynard) - Mark Sadang (https://github.com/msadang) URL: #585
closes rapidsai/cugraph#2237 Authors: - Divye Gala (https://github.com/divyegala) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #623
Changes: - Move `RngState` and `GeneratorType` to own header so they can be used in public APIs without a CUDA-compatible compiler (useful for python bindings etc.). - Move the type of generator into the state object - Add the `advance` method to the state object as well, allowing libraries to use `RngState` for host-only code - Remove the `rng.hpp` header: if we want to expose the `Rng` object itself as a host-only object, the implementation/definition would have to be separated from the declaration, and this means it would be hard to keep a header-only library for these objects. Otherwise, there is simply no way that this `rng.hpp` actually fulfills its contract of a host-only header, so we remove it. This is not the only way of creating a host-only API, but it's most likely the one which needs the least amount of changes in dependent libraries, and these libraries can basically "opt-in" to the host-only API. @cjnolet Do you agree with this approach? I'd like to get this PR into 22.06 fairly soon, so we can finally go back to sane APIs and compile commands in cugraph-ops and cugraph. Authors: - Matt Joux (https://github.com/MatthiasKohl) - Corey J. Nolet (https://github.com/cjnolet) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #609
Fixing changelog merge conflict between 22.04 and 22.06 Authors: - Corey J. Nolet (https://github.com/cjnolet) - Ray Douglass (https://github.com/raydouglass) Approvers: - Mark Sadang (https://github.com/msadang) URL: #625
After merging PR #609 , I realized that there were still a few bugs (all minor): 1. The deprecated `Rng` class had the `type` attribute removed. I added it back. This was used in cugraph-ops but not in other projects (cugraph, cuml). 2. The template function to call another function with the device state was actually unusable since you wouldn't be able to instantiate a function to call it with. So I removed that. 3. The macro used to call functions with a device state wasn't using fully namespace-qualified names, which made it unusable outside of RAFT. 4. Removed the deprecation warning of the new `rng_device.cuh` since it has to be included anyway through `rng.cuh`, so those deprecations didn't make sense. Authors: - Matt Joux (https://github.com/MatthiasKohl) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #630
* Derive the offset_policy from the mixin instead of the default accessor. Partially address #501 . Authors: - Jiaming Yuan (https://github.com/trivialfis) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #628
This PR fixes a CMake configuration failure when using RAFT's build dir as the package root, e.g.:
```
cmake -S /cuml/cpp -B /cuml/cpp/build -Draft_ROOT=/raft/cpp/build
```
CMake fails with this mysterious error:
```shell
The link interface of target "raft::raft" contains:
std::mdspan
but the target was not found. Possible reasons include:
* There is a typo in the target name.
* A find_package call is missing for an IMPORTED target.
* An ALIAS target is missing.
```
However the problem isn't `mdspan`.
<details><summary>The error occurs when raft-dependencies.cmake tries and fails to find libcudacxx:</summary><pre>
raft/cpp/build/release/raft-dependencies.cmake(206): find_dependency(libcudacxx )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(34): set(cmake_fd_quiet_arg )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(35): if(raft_FIND_QUIETLY )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(36): set(cmake_fd_quiet_arg QUIET )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(38): set(cmake_fd_required_arg )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(39): if(raft_FIND_REQUIRED )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(43): get_property(cmake_fd_alreadyTransitive GLOBAL PROPERTY _CMAKE_libcudacxx_TRANSITIVE_DEPENDENCY )
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake(47): find_package(libcudacxx QUIET )
CMake Debug Log at conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/CMakeFindDependencyMacro.cmake:47 (find_package):
find_package considered the following paths for Findlibcudacxx.cmake:<br/>
cuml/cpp/build/cuda-11.6.0/branch-22.06/release/CPM_modules/Findlibcudacxx.cmake
cuml/cpp/build/cuda-11.6.0/branch-22.06/release/_deps/rapids-cmake-src/rapids-cmake/Findlibcudacxx.cmake
conda/cuda_11.6/envs/rapids/share/cmake-3.23/Modules/Findlibcudacxx.cmake<br/>
The file was not found.<br/>
<PackageName>_ROOT CMake variable [CMAKE_FIND_USE_PACKAGE_ROOT_PATH].<br/>
raft/cpp/build/release<br/>
CMAKE_PREFIX_PATH variable [CMAKE_FIND_USE_CMAKE_PATH].<br/>
conda/cuda_11.6/envs/rapids<br/>
CMAKE_FRAMEWORK_PATH and CMAKE_APPBUNDLE_PATH variables
[CMAKE_FIND_USE_CMAKE_PATH].<br/>
Env variable libcudacxx_DIR [CMAKE_FIND_USE_CMAKE_ENVIRONMENT_PATH].<br/>
none<br/>
CMAKE_PREFIX_PATH env variable [CMAKE_FIND_USE_CMAKE_ENVIRONMENT_PATH].<br/>
none<br/>
CMAKE_FRAMEWORK_PATH and CMAKE_APPBUNDLE_PATH env variables
[CMAKE_FIND_USE_CMAKE_ENVIRONMENT_PATH].<br/>
none<br/>
Paths specified by the find_package HINTS option.<br/>
none<br/>
Standard system environment variables
[CMAKE_FIND_USE_SYSTEM_ENVIRONMENT_PATH].<br/>
conda/cuda_11.6
/usr/local
/usr
/
/usr/local/cuda<br/>
CMake User Package Registry [CMAKE_FIND_USE_PACKAGE_REGISTRY].<br/>
none<br/>
CMake variables defined in the Platform file
[CMAKE_FIND_USE_CMAKE_SYSTEM_PATH].<br/>
/usr/X11R6
/usr/pkg
/opt<br/>
CMake System Package Registry
[CMAKE_FIND_PACKAGE_NO_SYSTEM_PACKAGE_REGISTRY].<br/>
none<br/>
Paths specified by the find_package PATHS option.<br/>
none<br/>
find_package considered the following locations for libcudacxx's Config
module:<br/>
raft/cpp/build/release/libcudacxxConfig.cmake
raft/cpp/build/release/libcudacxx-config.cmake
raft/cpp/build/release/cmake/libcudacxxConfig.cmake
raft/cpp/build/release/cmake/libcudacxx-config.cmake
conda/cuda_11.6/envs/rapids/libcudacxxConfig.cmake
conda/cuda_11.6/envs/rapids/libcudacxx-config.cmake
conda/cuda_11.6/libcudacxxConfig.cmake
conda/cuda_11.6/libcudacxx-config.cmake
/usr/local/libcudacxxConfig.cmake
/usr/local/libcudacxx-config.cmake
/usr/libcudacxxConfig.cmake
/usr/libcudacxx-config.cmake
/libcudacxxConfig.cmake
/libcudacxx-config.cmake
/usr/local/cuda/libcudacxxConfig.cmake
/usr/local/cuda/libcudacxx-config.cmake
/opt/libcudacxxConfig.cmake
/opt/libcudacxx-config.cmake<br/>
The file was not found.<br/>
Call Stack (most recent call first):
raft/cpp/build/release/raft-dependencies.cmake:206 (find_dependency)
raft/cpp/build/release/raft-config.cmake:85 (include)
build/cuda-11.6.0/branch-22.06/release/cmake/CPM_0.35.0.cmake:215 (find_package)
build/cuda-11.6.0/branch-22.06/release/cmake/CPM_0.35.0.cmake:272 (cpm_find_package)
build/cuda-11.6.0/branch-22.06/release/_deps/rapids-cmake-src/rapids-cmake/cpm/find.cmake:152 (CPMFindPackage)
cmake/thirdparty/get_raft.cmake:50 (rapids_cpm_find)
cmake/thirdparty/get_raft.cmake:80 (find_and_configure_raft)
CMakeLists.txt:200 (include)
</pre></details>
The cause of the failure is this:
1. RAFT's `CMakeLists.txt` adds `cuco` before adding `libcudacxx` ([here](https://github.com/rapidsai/raft/blob/f487da7f2585e60f0a11aef43ce17c990cc6145a/cpp/CMakeLists.txt#L145-L146))
2. `cuco` is added to the `raft-distance-exports` export set ([here](https://github.com/rapidsai/raft/blob/f487da7f2585e60f0a11aef43ce17c990cc6145a/cpp/cmake/thirdparty/get_cuco.cmake#L21-L22))
* _cuco adds `libcudacxx`_, because it was declared before `libcudacxx` in step 1:
```
-- CPM: cuco: adding package [email protected] (1.7.0)
```
3. `libcudacxx` is added to the `raft-exports` export set ([here](https://github.com/rapidsai/raft/blob/f487da7f2585e60f0a11aef43ce17c990cc6145a/cpp/cmake/thirdparty/get_libcudacxx.cmake#L20-L21))
* Since `libcudacxx` has already been added by `cuco` in step 2, a `find_dependency(libcudacxx)` call (not a `CPMFindPackage` call) is recorded in `raft-dependencies.cmake`
When someone uses the raft build dir as the package root, CMake runs `raft-dependencies.cmake` before `raft-distance-dependencies.cmake`. `raft-dependencies.cmake` fails at `find_dependency(libcudacxx)`, because the `CPMFindPackage("NAME;cuco;...)` call that is expected to introduce `libcudacxx` hasn't been executed yet.
## Alternative fix
There are two other possible fixes aside from the one in this PR:
1. Swap the order of [these two lines](https://github.com/rapidsai/raft/blob/f487da7f2585e60f0a11aef43ce17c990cc6145a/cpp/CMakeLists.txt#L145-L146) in `CMakeLists.txt`.
2. Change the [`libcudacxx` `BUILD_EXPORT_SET` and `INSTALL_EXPORT_SET`](https://github.com/rapidsai/raft/blob/f487da7f2585e60f0a11aef43ce17c990cc6145a/cpp/cmake/thirdparty/get_libcudacxx.cmake#L20-L21) to `raft-distance-exports` so `find_dependency(libcudacxx)` is recorded in `raft-distance-dependencies.cmake` after `cuco` (and cuco's `libcudacxx` dependency) has been found.
I didn't see any references to `<cuda/std/...>` in RAFT itself, so the cleanest option is to delete RAFT's `get_libcudacxx.cmake`.
Authors:
- Paul Taylor (https://github.com/trxcllnt)
Approvers:
- Robert Maynard (https://github.com/robertmaynard)
- Corey J. Nolet (https://github.com/cjnolet)
URL: #632
…ust`/`spdlog` locations via envvars in `setup.py` (#624) This PR adds `cuda_lib_dir` to `library_dirs`, and allows changing the `UCX`/`RMM`/`Thrust`/`spdlog` lib/include dirs via envvars. This enables running `setup.py` when these libraries aren't in the default locations (like `/usr/local` vs. `/usr`), or when they're not installed globally. Example: ```shell env UCX_HOME=/usr/local/ucx \ RMM_INCLUDE_DIR=/opt/rapids/rmm/include \ THRUST_INCLUDE_DIR=/opt/rapids/rmm/build/_deps/thrust-src \ SPDLOG_INCLUDE_DIR=/opt/rapids/rmm/build/_deps/spdlog-src/include \ python setup.py ... ``` Authors: - Paul Taylor (https://github.com/trxcllnt) Approvers: - Corey J. Nolet (https://github.com/cjnolet) - Sevag Hanssian (https://github.com/sevagh) URL: #624
Authors: - Divye Gala (https://github.com/divyegala) Approvers: - Corey J. Nolet (https://github.com/cjnolet) - William Hicks (https://github.com/wphicks) - Jiaming Yuan (https://github.com/trivialfis) URL: #601
Authors: - Corey J. Nolet (https://github.com/cjnolet) Approvers: - Dante Gama Dessavre (https://github.com/dantegd) URL: #634
…#639) This PR removes a cyclic dependency from `raft/random/rng_device.cuh` which can create compiler errors (only in some cases). Since there are no direct includes of `raft/random/rng_device.cuh` in cuml and cugraph, the impact of the PR is low. However, it will improve things for cugraph-ops and cuopt which use this header directly. Authors: - Matt Joux (https://github.com/MatthiasKohl) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #639
This extends #648 and maps the CMake install components to capture what I think is the desire to have them map to conda packages. Mapping: - CMake 'raft' componet => conda raft-headers - CMake 'raft_nn' componet => conda libraft-component - CMake 'raft_distance' componet => conda libraft-distance Authors: - Robert Maynard (https://github.com/robertmaynard) Approvers: - Jordan Jacobelli (https://github.com/Ethyling) - Corey J. Nolet (https://github.com/cjnolet) URL: #653
A few simplifications and tricks to improve the performance of the kernel: - Promote some constants to static constexpr - Allow `capacity < WarpSize` - Reduce the frequency of `sort` operations for `filtered` version - Remove `warp_sort::load` to simplify the api and implementation Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Tamas Bela Feher (https://github.com/tfeher) URL: #606
Corrects the issues we are seeing in CI with failed `faiss:faiss` dependency Authors: - Robert Maynard (https://github.com/robertmaynard) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #655
Allow passing `nullptr` as the input-indices-argument of `select_k`. This would imply the input indices are a range from `0` to `input_len - 1` and improve the performance by doing less global memory reads. Also makes input pointers `const` as they should be. Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Tamas Bela Feher (https://github.com/tfeher) URL: #618
We have a few folks out of the office and cuml's CI is breaking because of a couple subtle recent changes to RAFT's cmake infrastructure and how the resulting exports are being packaged up in the conda packages. I'm reverting these commits to free up cuml's CI until we figure out how to fix the issue. Authors: - Corey J. Nolet (https://github.com/cjnolet) Approvers: - William Hicks (https://github.com/wphicks) URL: #657
This PR changes the pyraft build system to use scikit-build and CMake. It depends on the changes in #633 and should be merged after that PR. This PR is largely complete, except that I am not yet sure exactly which libraries/headers are entirely necessary for the build. This version builds locally for me, but not all tests pass since I don't have nccl. However, I am a bit surprised to see the build succeeding without me even searching for UCX (perhaps this is because it's already on my system include path) so I need to figure out what do with UCX, NCCL, and the various CUDA math libraries (cusparse, cublas, etc). Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Sevag Hanssian (https://github.com/sevagh) - Ray Douglass (https://github.com/raydouglass) - Corey J. Nolet (https://github.com/cjnolet) URL: #644
If the dists (keys) are the same, different corresponding ids may end up in the selection due to non-deterministic nature of some implementations. This PR updates the comparison functor in the tests to account for this problem. Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #659
Fix [inout] parameter C being mistakenly set as const in one of the raft::gemm overloads . Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #664
This PR fixes a number of small errors that slipped through in #644 and #633. None of them materially affect the build (whether a preexisting libraft exists or not) since 1) benchmarks are currently turned off by default anyway in the C++ build, and 2) pyraft only depends on libraft for headers, not for compiled libraries. Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #662
A few fixes/improvements to the logger: - Removed one of the two duplicate versions (`common/detail/logger.hpp`) - Added `RAFT_LOG_TRACE_VEC` to trace device/host arrays - Added `inline` annotations to some of the helper functions (to fix the linker error) - Replaced `char[]` with `vector<char>` where applicable for safety and to address `clang-tidy` warnings - Changed the default logger level to be controlled by the macro `RAFT_ACTIVE_LEVEL` instead of the fixed `RAFT_LEVEL_INFO`, but set the default `RAFT_ACTIVE_LEVEL` to `RAFT_LEVEL_INFO`. Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Micka (https://github.com/lowener) - Tamas Bela Feher (https://github.com/tfeher) - Corey J. Nolet (https://github.com/cjnolet) URL: #673
Pinnings to be in-line with rapidsai/cudf#10965 Authors: - GALI PREM SAGAR (https://github.com/galipremsagar) Approvers: - Corey J. Nolet (https://github.com/cjnolet) - Ray Douglass (https://github.com/raydouglass) URL: #680
-- fix race in fusedL2knn smem read/write by adding a syncwarp reported by racecheck tool of compute-sanitizer. -- this addresses issue - #676 Authors: - Mahesh Doijade (https://github.com/mdoijade) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #679
Authors: - Corey J. Nolet (https://github.com/cjnolet) Approvers: - Brad Rees (https://github.com/BradReesWork) URL: #683
The specialization of `std::vector` when `T=bool` is unfortunately causing compilation issue in cuml because the `data()` function member is not implemented. And the elements may not be stored contiguously. (Link to the CI failure: https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci/job/cuml/job/prb/job/cuml-cpu-cuda-build-arm64/CUDA=11.5/1364/console) cc @achirkin Authors: - Micka (https://github.com/lowener) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #681
Make sure the gpu vector allocated for the data generation is deallocated before the other vectors are allocated. With this, the test uses under 13 GB of memory. Authors: - Artem M. Chirkin (https://github.com/achirkin) Approvers: - Corey J. Nolet (https://github.com/cjnolet) URL: #689
RAFT's RNG class provides methods that fill a buffer with random numbers belonging to various probability distribution functions. For example`uniform`, `normal` etc. In this methods, multiple instances of random number generator are used in parallel. Each cuda thread gets it own instance of generator. The instance is initialized with three values `seed`, `subsequence` and `offset`. The value for `seed` is common for all threads and thread id is used for `subsequence`. The `offset` is kept `0` for all instances. To fill the buffer, the threads work in grid strided fashion demonstrated by loop below:
```
for (size_t i = tid; i < buffer_length; i += total_number_of_threads) {
buffer[i] = rng.next();
}
```
Due to grid-striding of loops, consecutive elements in the buffer are `ith` element in consecutive subsequences. For example, 10th and 11th elements in the buffer would be 0th element in 10th and 11th subsequences. In the unit tests for cugraph, this scheme seems to introduce a slight bias in certain cases. Easy fix to the issue is to break the lock step increment of individual subsequences.
This PR sets different offset value for each subsequence by setting `offset = subsequence`. Note that this change has no effect on the period of the random number generator.
This should fix the cuGraph issue rapidsai/cugraph#2266 for now.
Authors:
- Vinay Deshpande (https://github.com/vinaydes)
Approvers:
- Thejaswi. N. S (https://github.com/teju85)
URL: #690
Pinnings to be inline with: rapidsai/cudf#10987 Authors: - GALI PREM SAGAR (https://github.com/galipremsagar) Approvers: - Sevag Hanssian (https://github.com/sevagh) URL: #693
…encies (#686) Authors: - Robert Maynard (https://github.com/robertmaynard) Approvers: - Vyas Ramasubramani (https://github.com/vyasr) URL: #686
Discussions with the team indicated a general preference for having build.sh default to finding the C++ library by default (and erroring if one is not found) rather than building a new copy of libraft when building raft Python. Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Corey J. Nolet (https://github.com/cjnolet) - AJ Schmidt (https://github.com/ajschmidt8) URL: #697
Our `devel` Docker containers need to be switched to using `conda` compilers to resolve a linking error. `raft` is in those containers, but hasn't yet been built with `conda` compilers. This PR addresses that. These changes won't cleanly merge into `branch-22.08` unfortunately due to the changes in #641, but we can address that another time. Authors: - AJ Schmidt (https://github.com/ajschmidt8) - Corey J. Nolet (https://github.com/cjnolet) - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Corey J. Nolet (https://github.com/cjnolet)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR serves as the mechanism to release v22.06 of raft