| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
add copyrignt | 7 个月前 | |
fisrt commit init code | 8 个月前 | |
fisrt commit init code | 8 个月前 | |
fix: multiple improvements for open deployment and IO thread efficiency api/python/setup.py: - Restore 3-package build structure: openyuanrong, openyuanrong-sdk, openyuanrong-cpp-sdk - Fix run_ext to call copy_openyuanrong_runtime (not copy_openyuanrong) for OPENYUANRONG package type build.sh: - Fix -P flag to build all 3 packages with correct SETUP_TYPE values src/libruntime: - libruntime_config.h: decouple httpIocThreadsNum from connection count; default to hardware_concurrency() instead of 200 - downgrade.cpp: set httpIocThreadsNum=hardware_concurrency(), maxConnSize=10 - gwclient: use a shared io_context pool (size=hardware_concurrency()) so one thread handles multiple async connections via boost::asio event loop functionsystem (submodule): - ckpt_file_manager_actor.cpp: use $HOME instead of /home/yuanrong - container_executor.cpp: use runtimeHomeDir instead of /home/yuanrong - install.sh: patch python executor code_path at startup; fix missing {functionSchedulerLeasePort} substitution test/smoke/minimal-python/run_smoke.sh: - Fix PROJECT_ROOT (../.. -> ../../..) - Add --runtime_home_dir ${HOME} to yr start - Unset http_proxy vars so curl reaches local cluster addresses directly Signed-off-by: yuchaow <wangyuchao12@huawei.com> | 2 个月前 | |
feat: separate datasystem node roles for sandbox deployments The sandbox deployment needs datasystem master pods to stay out of worker hash-ring traffic while worker pods still join the ring. Thread node_role through process config, the shared datasystem deploy script, CI container setup, and ST coverage, then carry the datasystem submodule fix and repo-local GitCode PR helper alignment in one reviewable branch commit. Constraint: PR targets upstream/feature/sandbox and must carry the datasystem submodule fix with the deployment wiring Constraint: K8s startup goes through yr start, so datasystem behavior must live in the packaged shared deploy/data_system/install.sh path Constraint: Repo-local PR helper should match the fork-to-upstream GitCode flow used from this checkout Rejected: Keep master and worker roles inferred only from deployment shape | explicit node_role avoids ambiguous behavior when distributed master is enabled Rejected: Override the packaged datasystem install script inside the K8s image | yr start already starts datasystem from the packaged shared deploy path Rejected: Leave local user/home assumptions in docker-compose.yml | CI compose defaults should follow COMPILE_USER/COMPILE_HOME or caller USER/HOME Confidence: high Scope-risk: moderate Reversibility: clean Directive: Do not collapse node_role back into enable_distributed_master without validating worker hash-ring membership Tested: git diff --check; docker compose -f ci/ubuntu/docker-compose.yml config; env -u COMPILE_HOME -u COMPILE_USER HOME=/home/builder USER=builder docker compose -f ci/ubuntu/docker-compose.yml config; python -m py_compile deploy/sandbox/k8s/tests/test_yr_k8s_layout.py test/st/python/test_node_role.py; python -m pytest deploy/sandbox/k8s/tests/test_yr_k8s_layout.py -q; bash -n test/st/python/run_node_role_test.sh deploy/data_system/install.sh deploy/process/config.sh deploy/sandbox/k8s/bin/start-master.sh deploy/sandbox/k8s/bin/start-node.sh deploy/sandbox/k8s/bin/start-frontend.sh Not-tested: full sandbox ST cluster execution Signed-off-by: yuchaow <wangyuchao12@huawei.com> | 2 个月前 | |
feat: separate datasystem node roles for sandbox deployments The sandbox deployment needs datasystem master pods to stay out of worker hash-ring traffic while worker pods still join the ring. Thread node_role through process config, the shared datasystem deploy script, CI container setup, and ST coverage, then carry the datasystem submodule fix and repo-local GitCode PR helper alignment in one reviewable branch commit. Constraint: PR targets upstream/feature/sandbox and must carry the datasystem submodule fix with the deployment wiring Constraint: K8s startup goes through yr start, so datasystem behavior must live in the packaged shared deploy/data_system/install.sh path Constraint: Repo-local PR helper should match the fork-to-upstream GitCode flow used from this checkout Rejected: Keep master and worker roles inferred only from deployment shape | explicit node_role avoids ambiguous behavior when distributed master is enabled Rejected: Override the packaged datasystem install script inside the K8s image | yr start already starts datasystem from the packaged shared deploy path Rejected: Leave local user/home assumptions in docker-compose.yml | CI compose defaults should follow COMPILE_USER/COMPILE_HOME or caller USER/HOME Confidence: high Scope-risk: moderate Reversibility: clean Directive: Do not collapse node_role back into enable_distributed_master without validating worker hash-ring membership Tested: git diff --check; docker compose -f ci/ubuntu/docker-compose.yml config; env -u COMPILE_HOME -u COMPILE_USER HOME=/home/builder USER=builder docker compose -f ci/ubuntu/docker-compose.yml config; python -m py_compile deploy/sandbox/k8s/tests/test_yr_k8s_layout.py test/st/python/test_node_role.py; python -m pytest deploy/sandbox/k8s/tests/test_yr_k8s_layout.py -q; bash -n test/st/python/run_node_role_test.sh deploy/data_system/install.sh deploy/process/config.sh deploy/sandbox/k8s/bin/start-master.sh deploy/sandbox/k8s/bin/start-node.sh deploy/sandbox/k8s/bin/start-frontend.sh Not-tested: full sandbox ST cluster execution Signed-off-by: yuchaow <wangyuchao12@huawei.com> | 2 个月前 | |
fix[runtime]: fix FiberPool maxConcurrency to use InvokeOptions.concurrency - Change InvokeOptions.concurrency from int=1 to Optional[int]=None - Only pass concurrency to customExtensions when explicitly set - Add GetConcurrencyFromOptions helper to determine default based on isAsync: - async task/actor: default 1000 - non-async task/actor: default 1 - Use user-specified concurrency for FiberPool creation instead of global config This fixes a bug where async method calls were using global max concurrency (1000) instead of the user-specified InvokeOptions.concurrency value, causing unexpected concurrent execution behavior. Signed-off-by: Claude Code <noreply@anthropic.com> Signed-off-by: For_YL <zhangtangwei@huawei.com> | 1 个月前 | |
fisrt commit init code | 8 个月前 | |
chore: consolidate sandbox runtime gate fixes Signed-off-by: mhsong2 <songminhui2@huawei.com> | 15 天前 | |
feat: separate datasystem node roles for sandbox deployments The sandbox deployment needs datasystem master pods to stay out of worker hash-ring traffic while worker pods still join the ring. Thread node_role through process config, the shared datasystem deploy script, CI container setup, and ST coverage, then carry the datasystem submodule fix and repo-local GitCode PR helper alignment in one reviewable branch commit. Constraint: PR targets upstream/feature/sandbox and must carry the datasystem submodule fix with the deployment wiring Constraint: K8s startup goes through yr start, so datasystem behavior must live in the packaged shared deploy/data_system/install.sh path Constraint: Repo-local PR helper should match the fork-to-upstream GitCode flow used from this checkout Rejected: Keep master and worker roles inferred only from deployment shape | explicit node_role avoids ambiguous behavior when distributed master is enabled Rejected: Override the packaged datasystem install script inside the K8s image | yr start already starts datasystem from the packaged shared deploy path Rejected: Leave local user/home assumptions in docker-compose.yml | CI compose defaults should follow COMPILE_USER/COMPILE_HOME or caller USER/HOME Confidence: high Scope-risk: moderate Reversibility: clean Directive: Do not collapse node_role back into enable_distributed_master without validating worker hash-ring membership Tested: git diff --check; docker compose -f ci/ubuntu/docker-compose.yml config; env -u COMPILE_HOME -u COMPILE_USER HOME=/home/builder USER=builder docker compose -f ci/ubuntu/docker-compose.yml config; python -m py_compile deploy/sandbox/k8s/tests/test_yr_k8s_layout.py test/st/python/test_node_role.py; python -m pytest deploy/sandbox/k8s/tests/test_yr_k8s_layout.py -q; bash -n test/st/python/run_node_role_test.sh deploy/data_system/install.sh deploy/process/config.sh deploy/sandbox/k8s/bin/start-master.sh deploy/sandbox/k8s/bin/start-node.sh deploy/sandbox/k8s/bin/start-frontend.sh Not-tested: full sandbox ST cluster execution Signed-off-by: yuchaow <wangyuchao12@huawei.com> | 2 个月前 | |
fix(st): verify reverse tunnel from sandbox Signed-off-by: mhsong2 <songminhui2@huawei.com> | 11 天前 | |
add copyrignt | 7 个月前 | |
fisrt commit init code | 8 个月前 | |
fix st and llt | 7 个月前 | |
fix(build): complete fixes for brpc/gflags adaptation + remove deprecated tests Fixes for datasystem PR #1122 (brpc dual-mode) integration: 1. bazel/datasystem_sdk.bzl: add libprotoc.so.25.5.0 to cc_library srcs. 2. api/java/BUILD.bazel: - Use real filename libgflags.so.2.2.2 in LIB_REQUIRED. - Add libprotoc.so.25.5.0 to LIB_REQUIRED. 3. api/java/.../jni/LoadUtil.java: - Add libgflags.so.2.2.2 and libprotoc.so.25.5.0 to EXTRACT_ONLY_LIBS. - Add libgflags.so.2.2.2 to SYMLINK_MAP. - Add libprotoc.so.25.5.0 to LOADING_SEQUENCE. 4. test: remove all deprecated MSetTx test cases (C++ 9 + Python 2 + Java 4 + Go 2). Signed-off-by: OuGongChang <ougongchang@huawei.com> | 12 天前 | |
chore: consolidate sandbox post-merge fixes Preserve the original sandbox merge history while folding the follow-up CI, codecheck, packaging, deploy, and conflict-resolution commits into one merge commit. The tree is identical to the previous MR head 25a7b4e1fd7ea62d90e2e310a9b7414e4fb2d8a4. Signed-off-by: mhsong2 <songminhui2@huawei.com> | 1 个月前 | |
fix st | 7 个月前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 7 个月前 | ||
| 8 个月前 | ||
| 8 个月前 | ||
| 2 个月前 | ||
| 2 个月前 | ||
| 2 个月前 | ||
| 1 个月前 | ||
| 8 个月前 | ||
| 15 天前 | ||
| 2 个月前 | ||
| 11 天前 | ||
| 7 个月前 | ||
| 8 个月前 | ||
| 7 个月前 | ||
| 12 天前 | ||
| 1 个月前 | ||
| 7 个月前 |