feat: make stage1 AMI build idempotent

The stage2 AMI build process is somewhat flaky, and retrying the job
if it fails results in an error as the stage1 AMI has already been
built. This change attempts to re-use the stage1 AMI if it can be
found. Further changes will be required to reduce the flakiness on the
stage2 build.

Author: darora

.github/workflows/ami-release-nix.yml

@@ -94,12 +94,7 @@ jobs:
           POSTGRES_MAJOR_VERSION: ${{ env.POSTGRES_MAJOR_VERSION }}
         run: |
           GIT_SHA=${{github.sha}}
-          nix run github:supabase/postgres/${GIT_SHA}#packer -- init ${{ matrix.arch.packer_template }}
-          nix run github:supabase/postgres/${GIT_SHA}#packer -- build -var "git-head-version=${GIT_SHA}" -var "packer-execution-id=${EXECUTION_ID}" -var-file="${{ matrix.arch.vars_file }}" -var-file="common-nix.vars.pkr.hcl" -var "ansible_arguments=-e postgresql_major=${POSTGRES_MAJOR_VERSION}" -var "region=us-east-1" -var 'ami_regions=["us-east-1"]' -var "ami_name=${{ matrix.arch.ami_name }}" ${{ matrix.arch.packer_template }}
 
-      - name: Find stage 1 AMI
-        run: |
-          GIT_SHA=${{github.sha}}
           PG_VERSION=$(sed -n 's/postgres-version = "\(.*\)"/\1/p' common-nix.vars.pkr.hcl)
           REGION="us-east-1"
 
@@ -117,12 +112,28 @@ jobs:
             --output text)
 
           if [ -z "$STAGE1_AMI_ID" ] || [ "$STAGE1_AMI_ID" = "None" ]; then
-            echo "ERROR: Failed to find stage 1 AMI"
-            exit 1
+            echo "Did not find stage1 AMI - building it"
+            nix run github:supabase/postgres/${GIT_SHA}#packer -- init ${{ matrix.arch.packer_template }}
+            nix run github:supabase/postgres/${GIT_SHA}#packer -- build -var "git-head-version=${GIT_SHA}" -var "packer-execution-id=${EXECUTION_ID}" -var-file="${{ matrix.arch.vars_file }}" -var-file="common-nix.vars.pkr.hcl" -var "ansible_arguments=-e postgresql_major=${POSTGRES_MAJOR_VERSION}" -var "region=us-east-1" -var 'ami_regions=["us-east-1"]' -var "ami_name=${{ matrix.arch.ami_name }}" ${{ matrix.arch.packer_template }}
+            STAGE1_AMI_ID=$(aws ec2 describe-images \
+              --region "$REGION" \
+              --owners self \
+              --filters \
+                "Name=tag:postgresVersion,Values=${PG_VERSION}-stage1" \
+                "Name=tag:sourceSha,Values=${GIT_SHA}" \
+                "Name=state,Values=available" \
+                "Name=architecture,Values=${{ matrix.arch.ami_arch_filter }}" \
+              --query 'Images[0].ImageId' \
+              --output text)
+            if [ -z "$STAGE1_AMI_ID" ] || [ "$STAGE1_AMI_ID" = "None" ]; then
+              echo "Failed to find stage1 AMI right after building it"
+              exit 1
+            fi
           fi
 
           echo "Found stage 1 AMI: $STAGE1_AMI_ID"
           echo "STAGE1_AMI_ID=$STAGE1_AMI_ID" >> "$GITHUB_ENV"
+          exit 0
 
       - name: Build AMI stage 2
         env: