Merge pull request #79 from nebius/SCHED-1210

SCHED-1210 use cuda_force_upgrade for upgrading CUDA version and upgrade ansible requirements.txt

Author: asteny

ansible/common-packages.yml

@@ -1,7 +1,9 @@
 ---
 
 - name: Install common packages
-  hosts: localhost
+  hosts:
+    - localhost
+    - all
   gather_facts: true
   gather_subset:
     - min

ansible/python.yml

@@ -1,7 +1,9 @@
 ---
 
 - name: Install python
-  hosts: localhost
+  hosts:
+    - localhost
+    - all
   become: true
   roles:
     - python

ansible/requirements.txt

@@ -1,36 +1,36 @@
-ansible==12.2.0
-ansible-compat==25.8.2
-ansible-core==2.19.4
-ansible-lint==25.9.2
+ansible==13.4.0
+ansible-compat==25.12.1
+ansible-core==2.20.4
+ansible-lint==26.3.0
 argcomplete==3.6.3
-attrs==25.4.0
-black==25.9.0
+attrs==26.1.0
+black==26.3.1
 bracex==2.6
 cffi==2.0.0
-charset-normalizer==3.4.4
-click==8.3.0
-cryptography==46.0.3
+charset-normalizer==3.4.6
+click==8.3.1
+cryptography==46.0.5
 distro==1.9.0
-filelock==3.20.1
-importlib_metadata==8.7.0
+filelock==3.25.2
+importlib_metadata==9.0.0
 Jinja2==3.1.6
-jsonschema==4.25.1
+jsonschema==4.26.0
 jsonschema-specifications==2025.9.1
 MarkupSafe==3.0.3
 mypy_extensions==1.1.0
-packaging==25.0
-pathspec==0.12.1
-platformdirs==4.5.0
-pycparser==2.23
-python-debian==1.0.1
-pytokens==0.2.0
+packaging==26.0
+pathspec==1.0.4
+platformdirs==4.9.4
+pycparser==3.0
+python-debian==1.1.0
+pytokens==0.4.1
 PyYAML==6.0.3
 referencing==0.37.0
 resolvelib==1.2.1
-rpds-py==0.28.0
-ruamel.yaml==0.18.16
-ruamel.yaml.clib==0.2.14
+rpds-py==0.30.0
+ruamel.yaml==0.19.1
+ruamel.yaml.clib==0.2.15
 subprocess-tee==0.4.2
 wcmatch==10.1
-yamllint==1.37.1
+yamllint==1.38.0
 zipp==3.23.0

ansible/roles/cuda/tasks/main.yml

@@ -54,24 +54,39 @@
   tags:
     - cuda
 
-- name: Unhold old CUDA packages
-  ansible.builtin.dpkg_selections:
-    name: "{{ item }}"
-    selection: install
-  loop: "{{ cuda_hold_packages[cuda_old_version] }}"
-  when: cuda_old_version is defined
+- name: Get held packages
+  ansible.builtin.command: apt-mark showhold
+  register: cuda_apt_hold
+  changed_when: false
+  check_mode: false
+  when: cuda_force_upgrade | default(false) | bool
+  tags:
+    - cuda
+
+- name: Unhold only held CUDA packages
+  ansible.builtin.command: "apt-mark unhold {{ item }}"
+  loop: >-
+    {{
+      cuda_apt_hold.stdout_lines
+      | intersect(cuda_hold_packages | dict2items | map(attribute='value') | flatten | unique)
+    }}
+  register: cuda_unhold_out
+  changed_when: "'Canceled hold on' in (unhold_out.stdout | default(''))"
+  failed_when: cuda_unhold_out.rc != 0
+  when: cuda_force_upgrade | default(false) | bool
   tags:
     - cuda
 
 - name: Remove old CUDA packages
   ansible.builtin.apt:
-    name: "{{ cuda_packages[cuda_old_version] }}"
+    name: "{{ cuda_hold_packages | dict2items | map(attribute='value') | flatten | unique }}"
     state: absent
     purge: true
     autoremove: true
     update_cache: false
     force_apt_get: true
-  when: cuda_old_version is defined
+    allow_change_held_packages: true
+  when: cuda_force_upgrade | default(false) | bool
   tags:
     - cuda
 

ansible/roles/dcgmi/tasks/main.yml

@@ -51,18 +51,6 @@
   tags:
     - dcgmi
 
-- name: Remove old dcgmi packages
-  ansible.builtin.apt:
-    name: "{{ dcgmi_cuda_packages_map[dcgmi_cuda_old_version] }}"
-    state: absent
-    purge: true
-    autoremove: true
-    update_cache: false
-    force_apt_get: true
-  when: dcgmi_cuda_old_version is defined
-  tags:
-    - dcgmi
-
 - name: Ensure dcgmi packages is installed for detected CUDA {{ dcgmi_cuda_major_version }}
   ansible.builtin.apt:
     name: "{{ dcgmi_packages }}"

ansible/slurm.yml

@@ -1,7 +1,9 @@
 ---
 
 - name: Install slurm packages
-  hosts: localhost
+  hosts:
+    - localhost
+    - all
   become: true
   gather_facts: true
   gather_subset: