Bring back documentation publishing to Canary builds

We are using the same approach as in
apache/airflow-site#1055 where we utilise
/mnt directory to store checked out airflow-site.

This requires to switch back to public runners for canary builds
as public runners have space available in /mnt

Fixes: #41931

.github/workflows/static-checks-mypy-docs.yml

@@ -236,8 +236,6 @@ jobs:
     timeout-minutes: 150
     name: "Publish documentation"
     needs: build-docs
-    # For canary runs we need to push documentation to AWS S3 and preparing it takes a lot of space
-    # So we should use self-hosted ASF runners for this
     runs-on: ${{ fromJSON(inputs.runs-on-as-json-docs-build) }}
     env:
       GITHUB_REPOSITORY: ${{ github.repository }}
@@ -248,9 +246,7 @@ jobs:
       INCLUDE_SUCCESS_OUTPUTS: "${{ inputs.include-success-outputs }}"
       PYTHON_MAJOR_MINOR_VERSION: "${{ inputs.default-python-version }}"
       VERBOSE: "true"
-      # Temporary disabled it until self-hosted ASF runners will be working again
-    if: false
-    # if: inputs.canary-run == 'true' && inputs.branch == 'main'
+    if: inputs.canary-run == 'true' && inputs.branch == 'main'
     steps:
       - name: "Cleanup repo"
         shell: bash
@@ -266,16 +262,22 @@ jobs:
         with:
           name: airflow-docs
           path: './docs/_build'
+      - name: Check disk space available
+        run: df -h
+      - name: Create /mnt/airflow-site directory
+        run: sudo mkdir -p /mnt/airflow-site && sudo chown -R "${USER}" /mnt/airflow-site
       - name: "Clone airflow-site"
         run: >
-          git clone https://github.com/apache/airflow-site.git ${GITHUB_WORKSPACE}/airflow-site &&
-          echo "AIRFLOW_SITE_DIRECTORY=${GITHUB_WORKSPACE}/airflow-site" >> "$GITHUB_ENV"
+          git clone https://github.com/apache/airflow-site.git /mnt/airflow-site/airflow-site &&
+          echo "AIRFLOW_SITE_DIRECTORY=/mnt/airflow-site/airflow-site" >> "$GITHUB_ENV"
       - name: "Prepare breeze & CI image: ${{ inputs.default-python-version }}:${{ inputs.image-tag }}"
         uses: ./.github/actions/prepare_breeze_and_image
       - name: "Publish docs"
         run: >
           breeze release-management publish-docs --override-versioned --run-in-parallel
           ${{ inputs.docs-list-as-string }}
+      - name: Check disk space available
+        run: df -h
       - name: "Generate back references for providers"
         run: breeze release-management add-back-references all-providers
       - name: "Generate back references for apache-airflow"
@@ -302,3 +304,4 @@ jobs:
       - name: "Upload documentation to AWS S3"
         if: inputs.branch == 'main'
         run: aws s3 sync --delete ./docs/_build s3://apache-airflow-docs
+        working-directory: /mnt/airflow-site/airflow-site

dev/breeze/src/airflow_breeze/utils/selective_checks.py

@@ -1228,10 +1228,11 @@ def runs_on_as_json_self_hosted_asf(self) -> str:
 
     @cached_property
     def runs_on_as_json_docs_build(self) -> str:
-        if self._is_canary_run():
-            return RUNS_ON_SELF_HOSTED_ASF_RUNNER
-        else:
-            return RUNS_ON_PUBLIC_RUNNER
+        # We used to run docs build on self-hosted runners because they had more space, but
+        # It turned out that public runners have a lot of space in /mnt folder that we can utilise
+        # but in the future we might want to switch back to self-hosted runners so we have this
+        # separate property to determine that and place to implement different logic if needed
+        return RUNS_ON_PUBLIC_RUNNER
 
     @cached_property
     def runs_on_as_json_public(self) -> str:

dev/breeze/tests/test_selective_checks.py

@@ -1909,11 +1909,8 @@ def test_helm_tests_trigger_ci_build(files: tuple[str, ...], expected_outputs: d
             "apache/airflow",
             (),
             dict(),
-            # TODO: revert it when we fix self-hosted runners
             '["ubuntu-22.04"]',
-            '["self-hosted", "asf-runner"]',
-            # '["self-hosted", "Linux", "X64"]',
-            # TODO: revert it when we fix self-hosted runners
+            '["ubuntu-22.04"]',
             "false",
             "false",
             # "true",