Dogfooding by acezen #2621
Replies: 10 comments 16 replies
This comment has been hidden.
This comment has been hidden.
This comment has been hidden.
This comment has been hidden.
-
building graph analytical enginegraphscope ➜ /workspaces/GraphScope (3b11b77f) $ ./gs make analytical
Setting the environment for development.
[WARNING] GRAPHSCOPE_HOME will set to source root (/workspaces/GraphScope) for development.
[WARNING] To use you assigned GRAPHSCOPE_HOME, export GRAPHSCOPE_ENV=prod.
Read the env: GRAPHSCOPE_HOME=/workspaces/GraphScope
# this file is located in 'src/make_command.sh'
# code for 'gs make' goes here
# you can edit it freely and regenerate (it will not be overwritten)
args:
- ${args[component]} = analytical
- ${args[--install-prefix]} = /opt/graphscope
Making component analytical
make: Nothing to be done for 'analytical'. reference: https://graphscope.io/docs/latest/analytical_engine/dev_and_test.html#build-analytical-engine |
Beta Was this translation helpful? Give feedback.
This comment has been hidden.
This comment has been hidden.
-
Can we unify the building process with ./gs make analytical
./gs make interactive
./gs make learning https://graphscope.io/docs/latest/analytical_engine/dev_and_test.html#build-analytical-engine |
Beta Was this translation helpful? Give feedback.
-
no document about how to install |
Beta Was this translation helpful? Give feedback.
This comment has been hidden.
This comment has been hidden.
-
the learning building command should be: ./gs make learning run with
|
Beta Was this translation helpful? Give feedback.
-
task2 - install gae, gie, glethere are no instruction to help how to install gie and gle |
Beta Was this translation helpful? Give feedback.
-
task2g = load_mordern_graph() got error:
|
Beta Was this translation helpful? Give feedback.
-
System info
Task 1: 在本地试用 GraphScope
ogbn_mag
内置图ogbn_mag
到一个简单图,选择点标签为paper
, 边标签为cites
, 属性都为空。ogbn_mag
图上启动交互引擎 (Interactive)ogbn_mag
图,使用 learning 定义并执行一个 GCN 训练过程,可参考 CI 或文档中定义的过程Task 2: 在本地开发调试 GraphScope
vineyardd
直接启动vineyard
grape_engine
, 通过grape_engine
直接运行 analytical_engine,然后退出 grape_engine 进程analytical_engine/core/server/analytical_server.cc
),并重新编译,启动并找到新添加的日志frontend-x-SNAPSHOT.jar
和gaia_executor
,直接运行这个 jar 包和这个 gaia_executor,使 frontend 能够连接到 executor。interactive_engine/frontend/src/main/java/com/alibaba/graphscope/frontend/Frontend.java
) 和 executor (interactive_engine/executor/assembly/v6d/src/bin/gaia_executor.rs
) 各添加一条日志,并重新编译,启动并找到新添加的日志libgraphlearn_shared.so
Task 3: 运行本地测试
test_min.py
。(Python 中的单元测试更像是集成测试,因为往往由多个引擎参与)python/graphscope/tests/minitest
下新增一个test 文件,在里面添加一个涉及多个引擎的测试,其中要包括载入一张 ogbn_mag 图,打印 Kcore 算法结果,通过 Gremlin 查询图的点边数量,执行 GCN 训练。Task 4: 在 Kubernetes 上使用 GraphScope (Python SDK)
Task 5: 在 Kubernetes 上使用 GraphScope (Helm)
Task 6: 独立部署
modern_graph
,运行 LPA 算法Task 7: 打包镜像与了解 Kubernetes 上运行的基本概念
analytical_engine/core/server/analytical_server.cc
) 并重新构建镜像,在不使用docker push
的情况下,使用新的镜像启动并在日志中可以找到新添加的日志 (使用 registry 和 tag 来选择使用哪个镜像)interactive_engine/frontend/src/main/java/com/alibaba/graphscope/frontend/Frontend.java
) 和 executor (interactive_engine/executor/assembly/v6d/src/bin/gaia_executor.rs
) 各添加一条日志 并重新构建 Frontend 和 Executor 镜像,在不使用docker push
的情况下,使用新的镜像启动并可以找到新添加的日志coordinator/gscoordinator/coordinator.py
),并重新构建 Coordinator 镜像。启动并找到异常的报错。(日志在控制台中或者在 Coordinator Pod 的日志中)Task 8: 运行 K8S 上的测试
test_demo_scripts.py
中的test_demo_distribute
测试Task 9 : 打包为 Wheel 包
Beta Was this translation helpful? Give feedback.
All reactions