[LAB-724] ansible 1.1.2 changes (#725)

Author: alabdao

infrastructure/ansible/files/compute.service

@@ -18,6 +18,8 @@ ExecStart=bacalhau serve \
 {% if ansible_ec2_instance_id is defined %}
   --labels instance-id={{ ansible_ec2_instance_id }} \
 {% endif %}
+  --job-selection-probe-http {{ receptor_url }} \
+  --max-job-execution-timeout "{{ bacalhau_max_job_execution_timeout | default('24h') }}" \
   --peer {{ requester_peer }} \
   --job-selection-accept-networked \
   --job-selection-data-locality anywhere

infrastructure/ansible/files/compute.yaml

@@ -10,9 +10,7 @@ ExecStart=bacalhau serve \
   --node-type requester \
   --ipfs-connect {{ ipfs_connect }} \
   --private-internal-ipfs=false \
-{% if receptor_url is defined %}
   --job-selection-probe-http {{ receptor_url }} \
-{% endif %}
   --labels owner={{ owner }} \
   --job-selection-accept-networked \
   --job-selection-data-locality anywhere \

infrastructure/ansible/files/requester.service

@@ -2,7 +2,7 @@
   remote_user: ubuntu
   hosts: "{{ target_hosts | default('tag_Type_compute') }}"
   vars:
-    nvidia_distribution: ubuntu2004
+    nvidia_distribution: ubuntu2204
     nvidia_container_toolkit_key_path: /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
     ipfs_path: /opt/local/ipfs
   environment:

infrastructure/ansible/provision_compute_instance.yaml

@@ -4,7 +4,7 @@
   hosts: "{{ target_hosts | default('tag_Type_compute_only') }}"
   gather_facts: true
   vars:
-    nvidia_distribution: ubuntu2004
+    nvidia_distribution: ubuntu2204
     ipfs_version: "0.18.0"
     ipfs_path: "/opt/ipfs"
     go_version: 1.20.3

infrastructure/ansible/provision_compute_only.yaml

@@ -48,19 +48,10 @@
     receptor_hostname: "receptor.{{ ansible_ec2_tags_instance_Env | lower }}.labdao.xyz"
   when: ansible_ec2_tags_instance_Env is defined
 
-# Bacalhau PeerID, example `curl -s bacalhau.staging.labdao.xyz:1234/node_info | jq -r '.PeerInfo.ID'`
-- name: Determine requester bacalhau peer id
-  ansible.builtin.uri:
-    url: "http://{{ requester_hostname }}:1234/node_info"
-    return_content: true
-  register: requester_output
-  when: requester_hostname is defined and ansible_ec2_tags_instance_Type is defined and ansible_ec2_tags_instance_Type is match("compute.*")
-
 - name: Set requester_peer url
   ansible.builtin.set_fact:
-    # requester_peer: /dns4/requester.staging.labdao.xyz/tcp/1235/p2p/QmeLa2fx2FMNDWbeY3UqjELc1MbKwNxggcmdBmLZepY6VK
-    requester_peer: "/dns4/{{ requester_hostname }}/tcp/1235/p2p/{{ requester_output.content | from_json | community.general.json_query('PeerInfo.ID') }}"
-  when: requester_output is defined and ansible_ec2_tags_instance_Type is defined and ansible_ec2_tags_instance_Type is match("compute.*")
+    requester_peer: "/dns4/{{ requester_hostname }}/tcp/1234/http"
+  when: ansible_ec2_tags_instance_Type is defined and ansible_ec2_tags_instance_Type is match("compute.*")
 
 - name: Set ipfs connect url
   ansible.builtin.set_fact:
@@ -97,16 +88,16 @@
     state: started
     enabled: true
 
-- name: Flush handler to ensure Bacalhau is running
-  ansible.builtin.meta: flush_handlers
-
-- name: Deploy config file
-  become: true
-  ansible.builtin.template:
-    src: "files/{{ bacalhau_node_type }}.yaml"
-    dest: /home/ubuntu/.bacalhau/config.yaml
-    owner: ubuntu
-    group: ubuntu
-    mode: "0644"
-  notify:
-    - Restart Bacalhau
+      # - name: Flush handler to ensure Bacalhau is running
+      #   ansible.builtin.meta: flush_handlers
+      #
+      # - name: Deploy config file
+      #   become: true
+      #   ansible.builtin.template:
+      #     src: "files/{{ bacalhau_node_type }}.yaml"
+      #     dest: /home/ubuntu/.bacalhau/config.yaml
+      #     owner: ubuntu
+      #     group: ubuntu
+      #     mode: "0644"
+      #   notify:
+      #     - Restart Bacalhau

infrastructure/ansible/tasks/install_bacalhau_tasks.yaml

@@ -3,19 +3,22 @@
     name: "{{ item }}"
     source: pull
   loop:
-    - "dfam/tetools:1.85"
-    - "docker.io/mcmenemy/plex-dlkcat@sha256:92cfec2676bc5b8d75ee37c0fc31eb9c2e43c73a2acce2e4aaa60066b7a56f38"
-    - "ghcr.io/labdao/diffdock:main@sha256:b00432de73478d3da578e4a16ee669178828109f3c7bf9c58d44bb7514f68629"
+    - "ubuntu"
+    - "alpine"
+    - "nvidia/cuda"
     - "ghcr.io/labdao/equibind:main@sha256:21a381d9ab1ff047565685044569c8536a55e489c9531326498b28d6b3cc244f"
     - "ghcr.io/labdao/equibind@sha256:ae2cec63b3924774727ed1c6c8af95cf4aaea2d3f0c5acbec56478505ccb2b07"
-    - "ghcr.io/labdao/oddt:main"
-    - "gnina/gnina:v1.0.1"
-    - "public.ecr.aws/p7l9w5o7/colabfold:latest"
-    - "public.ecr.aws/p7l9w5o7/protbert:latest"
-    - "public.ecr.aws/p7l9w5o7/rfdiffusion:latest@sha256:0a6ff53004958ee5e770b0b25cd7f270eaf9fc285f6e91f17ad4024d2cc4ea91"
-    - "quay.io/biocontainers/samtools@sha256:6f88956b747a67b2a39a3ff72c4de30e665239ee11db610624dd4298e30db1bf"
-    - "quay.io/labdao/blender@sha256:1ee7bdc2e100bc3a8c968025579be950f50e685b440bd610971248b168f2bb74"
-    - "quay.io/labdao/openbabel@sha256:1087315d7eda6d0632c9f9df72500ab9f6fef612c79bae7410473a2336f7be34"
-    - "staphb/fastqc:0.12.1@sha256:f5d8f72753269e0cee071fe198c89a59a1f8071445739b3398f7818f7cb039ae"
+      # - "dfam/tetools:1.85"
+      # - "docker.io/mcmenemy/plex-dlkcat@sha256:92cfec2676bc5b8d75ee37c0fc31eb9c2e43c73a2acce2e4aaa60066b7a56f38"
+      # - "ghcr.io/labdao/diffdock:main@sha256:b00432de73478d3da578e4a16ee669178828109f3c7bf9c58d44bb7514f68629"
+      # - "ghcr.io/labdao/oddt:main"
+      # - "gnina/gnina:v1.0.1"
+      # - "public.ecr.aws/p7l9w5o7/colabfold:latest"
+      # - "public.ecr.aws/p7l9w5o7/protbert:latest"
+      # - "public.ecr.aws/p7l9w5o7/rfdiffusion:latest@sha256:0a6ff53004958ee5e770b0b25cd7f270eaf9fc285f6e91f17ad4024d2cc4ea91"
+      # - "quay.io/biocontainers/samtools@sha256:6f88956b747a67b2a39a3ff72c4de30e665239ee11db610624dd4298e30db1bf"
+      # - "quay.io/labdao/blender@sha256:1ee7bdc2e100bc3a8c968025579be950f50e685b440bd610971248b168f2bb74"
+      # - "quay.io/labdao/openbabel@sha256:1087315d7eda6d0632c9f9df72500ab9f6fef612c79bae7410473a2336f7be34"
+      # - "staphb/fastqc:0.12.1@sha256:f5d8f72753269e0cee071fe198c89a59a1f8071445739b3398f7818f7cb039ae"
   async: 86400
   poll: 0