github workfows: allow credentials failure for runs with no secrets (#390)

Summary:
Pull Request resolved: #390

https://github.com/pytorch/torchx/runs/5191403515?check_suite_focus=true

{F701191215}

Our integration tests now fail when exporting from Meta or on external PRs because there's no secrets available to run the integration tests.

We used to continue and dryrun the tests but that behavior regressed in adcfcd7 this restores that continue behavior.

Reviewed By: kiukchung

Differential Revision: D34223966

fbshipit-source-id: 0f5bb62716f3bed9f933343508391031e0ff44fa

Author: d4l3k

.github/workflows/aws-batch-integration-tests.yaml

@@ -26,11 +26,14 @@ jobs:
           aws-region: us-west-2
           role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
           role-session-name: github-torchx
+        continue-on-error: true
       - name: Install dependencies
         run: |
           set -eux
           pip install -e .[dev]
       - name: Run AWS Batch Integration Tests
+        env:
+          AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}
         run: |
           set -ex
 

.github/workflows/components-integration-tests.yaml

@@ -26,6 +26,7 @@ jobs:
           aws-region: us-west-2
           role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
           role-session-name: github-torchx
+        continue-on-error: true
       - name: Configure Kube Config
         env:
           AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}

.github/workflows/kfp-integration-tests.yaml

@@ -27,6 +27,7 @@ jobs:
           aws-region: us-west-2
           role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
           role-session-name: github-torchx
+        continue-on-error: true
       - name: Configure Kube Config
         env:
           AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}

.github/workflows/kubernetes-dist-train-integration-tests.yaml

@@ -26,6 +26,7 @@ jobs:
           aws-region: us-west-2
           role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
           role-session-name: github-torchx
+        continue-on-error: true
       - name: Configure Kube Config
         env:
           AWS_ROLE_ARN: ${{ secrets.AWS_ROLE_ARN }}

.github/workflows/slurm-integration-tests.yaml

@@ -26,6 +26,7 @@ jobs:
           aws-region: us-west-2
           role-to-assume: ${{ secrets.AWS_ROLE_ARN }}
           role-session-name: github-torchx
+        continue-on-error: true
       - name: Install Dependencies
         run:
           set -ex

scripts/awsbatchint.sh

@@ -7,14 +7,19 @@
 
 set -ex
 
-APP_ID="$(torchx run --wait --scheduler aws_batch -c queue=torchx utils.echo)"
-torchx status "$APP_ID"
-torchx describe "$APP_ID"
-torchx log "$APP_ID"
-LINES="$(torchx log "$APP_ID" | wc -l)"
+if [ -z "$AWS_ROLE_ARN" ]; then
+  # only dryrun if no secrets
+  torchx run --wait --scheduler aws_batch --dryrun -c queue=torchx utils.echo
+else
+  APP_ID="$(torchx run --wait --scheduler aws_batch -c queue=torchx utils.echo)"
+  torchx status "$APP_ID"
+  torchx describe "$APP_ID"
+  torchx log "$APP_ID"
+  LINES="$(torchx log "$APP_ID" | wc -l)"
 
-if [ "$LINES" -ne 1 ]
-then
-    echo "expected 1 log lines"
-    exit 1
+  if [ "$LINES" -ne 1 ]
+  then
+      echo "expected 1 log lines"
+      exit 1
+  fi
 fi