Add download retries option (#9911)

Signed-off-by: tu1h <lihai.tu@daocloud.io>

roles/download/defaults/main.yml

@@ -50,6 +50,9 @@ download_validate_certs: true
 # Use the first kube_control_plane if download_localhost is not set
 download_delegate: "{% if download_localhost %}localhost{% else %}{{ groups['kube_control_plane'][0] }}{% endif %}"
 
+# Allow control the times of download retries for files and containers
+download_retries: 4
+
 # The docker_image_info_command might seems weird but we are using raw/endraw and `{{ `{{` }}` to manage the double jinja2 processing
 docker_image_pull_command: "{{ docker_bin_dir }}/docker pull"
 docker_image_info_command: "{{ docker_bin_dir }}/docker images -q | xargs -i {{ '{{' }} docker_bin_dir }}/docker inspect -f {% raw %}'{{ '{{' }} if .RepoTags }}{{ '{{' }} join .RepoTags \",\" }}{{ '{{' }} end }}{{ '{{' }} if .RepoDigests }},{{ '{{' }} join .RepoDigests \",\" }}{{ '{{' }} end }}' {% endraw %} {} | tr '\n' ','"

roles/download/tasks/download_container.yml

@@ -60,7 +60,7 @@
       register: pull_task_result
       until: pull_task_result is succeeded
       delay: "{{ retry_stagger | random + 3 }}"
-      retries: 4
+      retries: "{{ download_retries }}"
       become: "{{ user_can_become_root | default(false) or not download_localhost }}"
       environment: "{{ proxy_env if container_manager == 'containerd' else omit }}"
       when:
@@ -101,7 +101,7 @@
       register: upload_image
       failed_when: not upload_image
       until: upload_image is succeeded
-      retries: 4
+      retries: "{{ download_retries }}"
       delay: "{{ retry_stagger | random + 3 }}"
       when:
         - pull_required

roles/download/tasks/download_file.yml

@@ -64,7 +64,7 @@
     run_once: "{{ download_force_cache }}"
     register: uri_result
     until: uri_result is success
-    retries: 4
+    retries: "{{ download_retries }}"
     delay: "{{ retry_stagger | default(5) }}"
     environment: "{{ proxy_env }}"
     no_log: "{{ not (unsafe_show_logs|bool) }}"
@@ -97,7 +97,7 @@
     register: get_url_result
     become: "{{ not download_localhost }}"
     until: "'OK' in get_url_result.msg or 'file already exists' in get_url_result.msg"
-    retries: 4
+    retries: "{{ download_retries }}"
     delay: "{{ retry_stagger | default(5) }}"
     environment: "{{ proxy_env }}"
     no_log: "{{ not (unsafe_show_logs|bool) }}"
@@ -122,7 +122,7 @@
     register: get_task
     until: get_task is succeeded
     delay: "{{ retry_stagger | random + 3 }}"
-    retries: 4
+    retries: "{{ download_retries }}"
     when:
     - download_force_cache