Skip to content

GitLab

Unverified Commit 24c875c6 authored by Rhett Ying's avatar Rhett Ying Committed by GitHub
Browse files

[CI] add more nodes for linux jobs (#6172)

parent 89f8d5f6
Show whitespace changes
Inline Side-by-side
Showing with 63 additions and 18 deletions
Jenkinsfile
View file @ 24c875c6
#!/usr/bin/env groovy #!/usr/bin/env groovy
// CI tests are executed within Docker containers as the 'root' user. However,
// communications between Jenkins nodes are done with the 'ubuntu' user(login
// via root is disallowed on AWS EC2 instances). Therefore, we need to change
// the file permission to allow 'ubuntu' user to access the files created by
// the 'root' user. This is achieved by running 'chmod -R 777 .'.
// Summary of Jenkins nodes:
// - linux-benchmark-node: Linux CPU node for authentication and lint check.
// number of nodes: 1
// instance type: m5.2xlarge(8 vCPUs, 32 GB memory)
// number of executors per node: 6
// number of jobs running on this node per CI run: 3
// - linux-cpu-node: Linux CPU node for building and testing.
// number of nodes: 4
// instance type: m6i.24xlarge(96 vCPUs, 384 GB memory)
// number of executors per node: 6
// number of jobs running on this node per CI run: 8
// - linux-gpu-node: Linux GPU node for building and testing.
// number of nodes: 4
// instance type: g4dn.4xlarge(16 vCPUs, 64 GB memory, 1 GPU)
// number of executors per node: 1
// number of jobs running on this node per CI run: 4
// - windows-node: Windows CPU node for building and testing.
// number of nodes: 1
// instance type: m5.8xlarge(32 vCPUs, 128 GB memory)
// number of executors per node: 2
// number of jobs running on this node per CI run: 3
dgl_linux_libs = 'build/libdgl.so, build/runUnitTests, python/dgl/_ffi/_cy3/core.cpython-*-x86_64-linux-gnu.so, build/tensoradapter/pytorch/*.so, build/dgl_sparse/*.so, build/graphbolt/*.so' dgl_linux_libs = 'build/libdgl.so, build/runUnitTests, python/dgl/_ffi/_cy3/core.cpython-*-x86_64-linux-gnu.so, build/tensoradapter/pytorch/*.so, build/dgl_sparse/*.so, build/graphbolt/*.so'
// Currently DGL on Windows is not working with Cython yet // Currently DGL on Windows is not working with Cython yet
dgl_win64_libs = "build\\dgl.dll, build\\runUnitTests.exe, build\\tensoradapter\\pytorch\\*.dll, build\\dgl_sparse\\*.dll, build\\graphbolt\\*.dll" dgl_win64_libs = "build\\dgl.dll, build\\runUnitTests.exe, build\\tensoradapter\\pytorch\\*.dll, build\\dgl_sparse\\*.dll, build\\graphbolt\\*.dll"
def init_git() { def init_git() {
sh "chmod -R 777 ." // Fix permission issue
sh 'rm -rf *' sh 'rm -rf *'
sh "git config --global --add safe.directory '*'"
checkout scm checkout scm
sh 'git submodule update --recursive --init' sh 'git submodule update --recursive --init'
} }
...@@ -286,7 +316,7 @@ pipeline { ...@@ -286,7 +316,7 @@ pipeline {
stage('CPU Build') { stage('CPU Build') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230711" image "dgllib/dgl-ci-cpu:v230711"
args "-u root" args "-u root"
alwaysPull true alwaysPull true
...@@ -297,6 +327,7 @@ pipeline { ...@@ -297,6 +327,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -304,7 +335,7 @@ pipeline { ...@@ -304,7 +335,7 @@ pipeline {
stage('GPU Build') { stage('GPU Build') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-gpu:cu116_v230711" image "dgllib/dgl-ci-gpu:cu116_v230711"
args "-u root" args "-u root"
alwaysPull true alwaysPull true
...@@ -316,6 +347,7 @@ pipeline { ...@@ -316,6 +347,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -323,7 +355,7 @@ pipeline { ...@@ -323,7 +355,7 @@ pipeline {
stage('PyTorch Cugraph GPU Build') { stage('PyTorch Cugraph GPU Build') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "rapidsai/cugraph_stable_torch-cuda:11.8-base-ubuntu20.04-py3.10-pytorch2.0.0-rapids23.04" image "rapidsai/cugraph_stable_torch-cuda:11.8-base-ubuntu20.04-py3.10-pytorch2.0.0-rapids23.04"
args "-u root" args "-u root"
alwaysPull true alwaysPull true
...@@ -334,6 +366,7 @@ pipeline { ...@@ -334,6 +366,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -359,8 +392,9 @@ pipeline { ...@@ -359,8 +392,9 @@ pipeline {
stage('C++ CPU') { stage('C++ CPU') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230711" image "dgllib/dgl-ci-cpu:v230711"
args "-u root"
alwaysPull true alwaysPull true
} }
} }
...@@ -369,6 +403,7 @@ pipeline { ...@@ -369,6 +403,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -376,9 +411,9 @@ pipeline { ...@@ -376,9 +411,9 @@ pipeline {
stage('C++ GPU') { stage('C++ GPU') {
agent { agent {
docker { docker {
label "linux-gpu-node" label "dgl-ci-linux-gpu"
image "dgllib/dgl-ci-gpu:cu116_v230711" image "dgllib/dgl-ci-gpu:cu116_v230711"
args "--runtime nvidia" args "-u root --runtime nvidia"
alwaysPull true alwaysPull true
} }
} }
...@@ -387,6 +422,7 @@ pipeline { ...@@ -387,6 +422,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -405,8 +441,9 @@ pipeline { ...@@ -405,8 +441,9 @@ pipeline {
stage('Tensorflow CPU') { stage('Tensorflow CPU') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230810" image "dgllib/dgl-ci-cpu:v230810"
args "-u root"
alwaysPull true alwaysPull true
} }
} }
...@@ -419,6 +456,7 @@ pipeline { ...@@ -419,6 +456,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -426,9 +464,9 @@ pipeline { ...@@ -426,9 +464,9 @@ pipeline {
stage('Tensorflow GPU') { stage('Tensorflow GPU') {
agent { agent {
docker { docker {
label "linux-gpu-node" label "dgl-ci-linux-gpu"
image "dgllib/dgl-ci-gpu:cu116_v230711" image "dgllib/dgl-ci-gpu:cu116_v230711"
args "--runtime nvidia" args "-u root --runtime nvidia"
alwaysPull true alwaysPull true
} }
} }
...@@ -443,6 +481,7 @@ pipeline { ...@@ -443,6 +481,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -450,9 +489,9 @@ pipeline { ...@@ -450,9 +489,9 @@ pipeline {
stage('Torch CPU') { stage('Torch CPU') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230711" image "dgllib/dgl-ci-cpu:v230711"
args "--shm-size=4gb" args "-u root --shm-size=4gb"
alwaysPull true alwaysPull true
} }
} }
...@@ -475,6 +514,7 @@ pipeline { ...@@ -475,6 +514,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -502,9 +542,9 @@ pipeline { ...@@ -502,9 +542,9 @@ pipeline {
stage('Torch GPU') { stage('Torch GPU') {
agent { agent {
docker { docker {
label "linux-gpu-node" label "dgl-ci-linux-gpu"
image "dgllib/dgl-ci-gpu:cu116_v230711" image "dgllib/dgl-ci-gpu:cu116_v230711"
args "--runtime nvidia --shm-size=8gb" args "-u root --runtime nvidia --shm-size=8gb"
alwaysPull true alwaysPull true
} }
} }
...@@ -523,6 +563,7 @@ pipeline { ...@@ -523,6 +563,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -530,9 +571,9 @@ pipeline { ...@@ -530,9 +571,9 @@ pipeline {
stage('Distributed') { stage('Distributed') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230711" image "dgllib/dgl-ci-cpu:v230711"
args "--shm-size=4gb" args "-u root --shm-size=4gb"
alwaysPull true alwaysPull true
} }
} }
...@@ -546,6 +587,7 @@ pipeline { ...@@ -546,6 +587,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -553,9 +595,9 @@ pipeline { ...@@ -553,9 +595,9 @@ pipeline {
stage('PyTorch Cugraph GPU') { stage('PyTorch Cugraph GPU') {
agent { agent {
docker { docker {
label "linux-gpu-node" label "dgl-ci-linux-gpu"
image "rapidsai/cugraph_stable_torch-cuda:11.8-base-ubuntu20.04-py3.10-pytorch2.0.0-rapids23.04" image "rapidsai/cugraph_stable_torch-cuda:11.8-base-ubuntu20.04-py3.10-pytorch2.0.0-rapids23.04"
args "--runtime nvidia --shm-size=8gb" args "-u root --runtime nvidia --shm-size=8gb"
alwaysPull true alwaysPull true
} }
} }
...@@ -569,6 +611,7 @@ pipeline { ...@@ -569,6 +611,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
...@@ -576,8 +619,9 @@ pipeline { ...@@ -576,8 +619,9 @@ pipeline {
stage('DGL-Go') { stage('DGL-Go') {
agent { agent {
docker { docker {
label "linux-cpu-node" label "dgl-ci-linux-cpu"
image "dgllib/dgl-ci-cpu:v230711" image "dgllib/dgl-ci-cpu:v230711"
args "-u root"
alwaysPull true alwaysPull true
} }
} }
...@@ -590,6 +634,7 @@ pipeline { ...@@ -590,6 +634,7 @@ pipeline {
} }
post { post {
always { always {
sh "chmod -R 777 ." // Fix permission issue
cleanWs disableDeferredWipeout: true, deleteDirs: true cleanWs disableDeferredWipeout: true, deleteDirs: true
} }
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment