note the nektos/act vs gitea act_runner network issue

.gitea/workflows/tests.yaml

@@ -9,6 +9,17 @@ on:
     - cron: "6 */12 * * *"
 
 jobs:
+  # there is some undesirable behavior when running tests because nektos/act mimicks github actions.
+  # we are banned from github so we aren't using that. instead, we use gitea act_runner.
+  # github actions Job runner runs on docker 'host' network.
+  # gitea act_runner runs on a custom named bridge network. e.g. `GITEA-ACTIONS-TASK-974_WORKFLOW-ci_JOB-Tests-Checks-test_phoenix-network`
+  # confusing, right? It totally is!
+  # when we are working with gitea act_runner, we can use dns names, e.g. test-db.
+  # when we are working with nektos/act, we must use the IP address of the service container.
+  # this causes issues because nektos/act will not be able to find test-db:5432 but act_runner can.
+  # this means we have inconsistent behavior across testing environments, which is bad.
+  # our workaround is to use localhost on local dev, and the dns name on gitea act_runner.
+
   test_phoenix:
     env:
       SECRET_KEY_BASE: ${{ secrets.SECRET_KEY_BASE }}
@@ -37,14 +48,15 @@ jobs:
       pull-requests: write
 
     services:
-      db:
+      test-db:
         image: postgres:16
-        ports:
-          - 5433:5432
         env:
           POSTGRES_DB: ${{ vars.DB_NAME }}
           POSTGRES_USER: ${{ vars.DB_USER }}
           POSTGRES_PASSWORD: ${{ secrets.DB_PASS }}
+          PGUSER: ${{ vars.DB_USER }}
+        ports:
+          - 5433:5432
         options: >-
           --health-cmd pg_isready
           --health-interval 10s
@@ -98,6 +110,14 @@ jobs:
             user = "nobody"
 
     steps:
+      # I wasn't successful with this. Something about the docker network in which the container runs?
+      # - name: postgres service check
+      #   env:
+      #     PGPASSWORD: ${{ secrets.DB_PASS }}
+      #   uses: docker://postgres:16
+      #   with:
+      #     args: /usr/bin/pg_isready --host test-db --port 5432
+
       - name: Setup FFmpeg
         uses: FedericoCarboni/setup-ffmpeg@v3
 
@@ -122,14 +142,14 @@ jobs:
           elixir-version: "1.16.0-otp-26"
           otp-version: "26.0"
 
-      - name: Mix and build cache
+      # - name: Mix and build cache
-        uses: actions/cache@v4
+      #   uses: actions/cache@v4
-        with:
+      #   with:
-          path: |
+      #     path: |
-            deps
+      #       deps
-            _build
+      #       _build
-          key: ${{ runner.os }}-mix-${{ hashFiles('**/mix.lock') }}
+      #     key: ${{ runner.os }}-mix-${{ hashFiles('**/mix.lock') }}
-          restore-keys: ${{ runner.os }}-mix-
+      #     restore-keys: ${{ runner.os }}-mix-
 
       - name: Get dependencies
         run: mix deps.get

README.md

@@ -41,11 +41,13 @@ Use devbox helper script
     cat ./backups/your-backup.sql | docker exec -i postgres_db psql -U postgres
 
 
-## Metrics Notes
+## testing
-
-Keeping track of metrics we want to scrape using Prometheus
-
-
-
-
 
+there is some undesirable behavior when running tests because nektos/act mimicks github actions.
+we are banned from github so we aren't using that. instead, we use gitea act_runner.
+github actions Job runner runs on docker 'host' network.
+gitea act_runner runs on a custom named bridge network. e.g. `GITEA-ACTIONS-TASK-974_WORKFLOW-ci_JOB-Tests-Checks-test_phoenix-network`
+confusing, right? It totally is!
+when we are working with gitea act_runner, we can use dns names, e.g. test-db.
+when we are working with nektos/act, we must use the IP address of the service container.
+WTF there must be a better way