Merge pull request #98 from SCOREC/ac/perlmutter-actions

GPU Systems Testing

.github/workflows/systems.yml

@@ -0,0 +1,67 @@
+name: Systems
+on:
+  schedule:
+    # Sunday 1:30 AM
+    - cron: '30 1 * * 0' 
+
+concurrency:
+  group: systems-omega_h
+  cancel-in-progress: true
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    timeout-minutes: 30
+    strategy:
+      matrix:
+        machine: ["Perlmutter", "Frontier"]
+
+    steps:
+
+    - name: checkout omega_h
+      uses: actions/checkout@v4
+      with:
+        path: omega_h
+
+    - name: setup python
+      uses: actions/setup-python@v5
+      with:
+        python-version: '3.10'
+
+    - name: install packing
+      run: sudo apt install python3-packaging
+
+    - name: install globus
+      run: |
+          python -m ensurepip --upgrade --user
+          python -m pip install globus-compute-endpoint
+
+    - name: use globus
+      working-directory: omega_h/.github/workflows
+      env:
+        GLOBUS_ID: ${{ secrets.GLOBUS_COMPUTE_ID }}
+        GLOBUS_SECRET: ${{ secrets.GLOBUS_COMPUTE_SECRET }}
+      run: |
+        export GLOBUS_COMPUTE_CLIENT_ID="$GLOBUS_ID"
+        export GLOBUS_COMPUTE_CLIENT_SECRET="$GLOBUS_SECRET"
+        if [ ${{matrix.machine}} == Perlmutter ]; then TARGET_ENDPOINT=0dd4499a-8d76-4977-bae9-841e4bb2f616; fi
+        if [ ${{matrix.machine}} == Frontier ]; then TARGET_ENDPOINT=d625c6cf-de7a-4228-ac44-56247a642fe0; fi
+        python test_on_system.py ${{ github.sha }} $TARGET_ENDPOINT
+
+    - name: print build log
+      working-directory: omega_h/.github/workflows
+      run: cat omega_h-test-result/Build.log
+
+    - name: print test summary
+      working-directory: omega_h/.github/workflows
+      run: cat omega_h-test-result/TestSummary.log
+
+    - name: print test log
+      working-directory: omega_h/.github/workflows
+      run: cat omega_h-test-result/LastTest.log
+
+    - name: check failed test
+      working-directory: omega_h/.github/workflows
+      run: if grep "Failed" omega_h-test-result/TestSummary.log; then return 1; fi
+
+

.github/workflows/test_on_system.py

@@ -0,0 +1,48 @@
+# How to use
+# 1. Login to https://app.globus.org/settings/developers and copy a project app id and secret
+# 2. Use the id and secret to create and endpoint https://funcx.readthedocs.io/en/latest/sdk.html#client-credentials-with-clients
+#     $ export FUNCX_SDK_CLIENT_ID="b0500dab-ebd4-430f-b962-0c85bd43bdbb"
+#     $ export FUNCX_SDK_CLIENT_SECRET="ABCDEFGHIJKLMNOP0123456789="
+# 3. Set up an endpoint on the computer that will run the tests, using these instructions: https://funcx.readthedocs.io/en/latest/endpoints.html
+# 4. Create install-test.sh and run-test.sh on target computer
+
+from globus_compute_sdk import Executor
+import sys
+import os
+
+name = "omega_h-test"
+build = "build-omega_h"
+branch = sys.argv[1]
+endpoint = sys.argv[2]
+
+def run_on_endpoint(name, build, branch):
+    import subprocess
+
+    install = subprocess.run(["./install-test.sh "+name+" "+branch], shell=True, encoding="utf_8", stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+    if install.returncode == 1:
+        return (install, None, None)
+
+    summary = subprocess.run(["./run-test.sh "+name+" "+build], shell=True, encoding="utf_8", stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
+    if summary.returncode == 1:
+        return (install, summary, None)
+
+    with open(name+"-result/LastTest.log","r")  as f:
+        result = f.read()
+
+    return (install, summary, result)
+
+gce = Executor(endpoint_id = endpoint)
+future = gce.submit(run_on_endpoint, name, build, branch)
+result = future.result()
+
+os.popen("mkdir -p "+name+"-result").read()
+with open(name+"-result/Build.log", "w") as text_file:
+    text_file.write("%s" % result[0].stdout)
+    text_file.close()
+if result[0].returncode == 0:
+    with open(name+"-result/TestSummary.log", "w") as text_file:
+        text_file.write("%s" % result[1].stdout)
+        text_file.close()
+    with open(name+"-result/LastTest.log", "w") as text_file:
+        text_file.write("%s" % result[2])
+        text_file.close()