WIP: cobald container containing and using slurm

inv.yml

@@ -19,6 +19,7 @@ all:
     slurm:
       hosts:
         ed-c7-1:
+        ed-c7-2:
     cobald:
       hosts:
         ed-c7-1:

play.yml

@@ -1,6 +1,8 @@
 ---
 - hosts: all
-  vars_files: ['vars-auth.yml']
+  vars_files:
+  - vars-auth.yml
+  - vars-influx.yml
   tasks:
   - name: "install epel repo" # for htop etc.
     yum:
@@ -53,27 +55,34 @@
       - htcondor-containered
       - htcondor
 
-- hosts: slurm
+- hosts: slurm, cobald
   vars:
     container_privileged: True
-    num_nodes: 3
+    slurm_num_nodes: 10
   roles:
   - name: "setup docker"
     role: docker
     tags: docker
+  - name: "get facts from existing cobald instance (i.e. hostname)"
+    role: cobald_facts
+    vars:
+      container_name: cobald
+    tags: [ slurm, cobald ]
   - name: "setup slurm test environment in docker containers"
     role: slurm
     vars:
       slurm_user: slurm # or root
+      num_nodes: "{{slurm_num_nodes}}"
+      extra_nodes:
+      - "{{cobald_container_hostname}}" # from cobald_facts, read or generated
+      docker_network: slurm
+    when: '"slurm" in group_names'
     tags: slurm
-
-- hosts: cobald
-  roles:
-  - name: "setup docker"
-    role: docker
-    tags: docker
   - name: "install cobald"
     role: cobald
     vars:
-      docker_network: slurm
+      cobald_slurm: True
+      container_name: cobald
+      # docker_network: slurm # overriden by vars/slurm.yml
+    when: '"cobald" in group_names'
     tags: cobald

roles/cobald/files/cobald-entrypoint.sh

@@ -1,7 +1,7 @@
 #!/bin/sh
 
-[ -f /usr/local/lib/cobaldmodules/setup.py -a \
+for i in /usr/local/lib/entrypoints.d/* ; do
-	-d /usr/local/lib/cobaldmodules/cobaldmodules ] && \
+    [ -f $i ] && /bin/sh $i || break
-		pip3 install --no-deps --editable /usr/local/lib/cobaldmodules
+done
 
 exec "${@:-/bin/bash}"

roles/cobald/files/grafana-dashboard.json

@@ -16,7 +16,7 @@
   "gnetId": null,
   "graphTooltip": 0,
   "id": 1,
-  "iteration": 1623184036048,
+  "iteration": 1623317629899,
   "links": [],
   "panels": [
     {
@@ -99,7 +99,7 @@
           ],
           "orderByTime": "ASC",
           "policy": "default",
-          "query": "nodes = from(bucket: \"batleth\")\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> keep(columns: [\"tardis_machine_name\", \"_time\"])\n  |> sort(columns: [\"_time\"], desc: true)\n  |> unique(column: \"tardis_machine_name\")\n//  |> yield()\n\noffset = from(bucket: \"batleth\")\n  |> range(start: 0, stop: v.timeRangeStart)\n  |> filter(fn: (r) => r._field == \"state\")\n  |> group(columns: [\"tardis_machine_name\", \"machine_type\"])\n  |> reduce(fn: (r, accumulator) => ({\n    _value: accumulator._value + (\n      if r._value == \"AvailableState\" then 1 \n      else if r._value == \"DownState\" then -1 \n      else 0)\n    }), identity: {_value: 0})\n  |> duplicate(column: \"_stop\", as: \"_time\")\n\nnew = from(bucket: \"batleth\")\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> filter(fn: (r) => r._field == \"state\")\n//  |> filter(fn: (r) => r.tardis_machine_name == \"${machine}\")\n  |> group(columns: [\"tardis_machine_name\", \"machine_type\"])\n  |> window(every: $__interval)\n  |> reduce(fn: (r, accumulator) => ({\n      _value: accumulator._value + (\n        if r._value == \"AvailableState\" then 1 \n        else if r._value == \"DownState\" then -1 \n        else 0)\n    }), identity: {_value: 0})\n  |> duplicate(column: \"_stop\", as: \"_time\")\n\nunion(tables: [offset, new])\n  |> window(every: inf)\n  |> cumulativeSum()\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> yield()",
+          "query": "f_r = (r, accumulator) => ({\n    _value: accumulator._value + (\n      if r._value == \"AvailableState\" then 1\n      else if r._value == \"DownState\" then -1\n      else 0)\n    })\n\nnodes = from(bucket: \"batleth\")\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> keep(columns: [\"tardis_machine_name\", \"_time\"])\n  |> sort(columns: [\"_time\"], desc: true)\n  |> unique(column: \"tardis_machine_name\")\n//  |> yield()\n\noffset = from(bucket: \"batleth\")\n  |> range(start: 0, stop: v.timeRangeStart)\n  |> filter(fn: (r) => r._field == \"state\")\n  |> group(columns: [\"tardis_machine_name\", \"machine_type\"])\n  |> reduce(fn: f_r, identity: {_value: 0})\n  |> duplicate(column: \"_stop\", as: \"_time\")\n\nnew = from(bucket: \"batleth\")\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> filter(fn: (r) => r._field == \"state\")\n//  |> filter(fn: (r) => r.tardis_machine_name == \"${machine}\")\n  |> group(columns: [\"tardis_machine_name\", \"machine_type\"])\n  |> window(every: $__interval)\n  |> reduce(fn: f_r, identity: {_value: 0})\n  |> duplicate(column: \"_stop\", as: \"_time\")\n\nunion(tables: [offset, new])\n  |> window(every: inf)\n  |> cumulativeSum()\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> yield()\n",
           "refId": "A",
           "resultFormat": "time_series",
           "select": [
@@ -550,10 +550,9 @@
       {
         "allValue": null,
         "current": {
-          "isNone": true,
           "selected": false,
-          "text": "None",
+          "text": "cobald-xvmcqc",
-          "value": ""
+          "value": "cobald-xvmcqc"
         },
         "datasource": "InfluxDB",
         "definition": "from(bucket: \"batleth\")\n  |> range(start: v.timeRangeStart, stop: v.timeRangeStop)\n  |> keep(columns: [\"tardis_machine_name\", \"_time\"])\n  |> sort(columns: [\"_time\"], desc: true)\n  |> unique(column: \"tardis_machine_name\")",
@@ -586,5 +585,5 @@
   "timezone": "",
   "title": "cobald",
   "uid": "urDuvE6Gk",
-  "version": 9
+  "version": 2
 }

roles/cobald/files/init-cobaldmodules.sh

@@ -0,0 +1,5 @@
+#!/bin/sh
+
+[ -f /usr/local/lib/cobaldmodules/setup.py -a \
+	-d /usr/local/lib/cobaldmodules/cobaldmodules ] && \
+		pip3 install --no-deps --editable /usr/local/lib/cobaldmodules

roles/cobald/tasks/main.yml

@@ -1,33 +1,41 @@
+- include_vars: cobald-slurm.yml
+  when: cobald_slurm | default(False)
+
 - file:
-    path: "/container/{{item}}/cobald/"
+    path: "/container/{{item.name}}/cobald{{item.pfx|default('')}}/"
     state: directory
     owner: "{{unpriv_user}}"
     group: docker
   loop:
-  - docker-images
+  - name: docker-images
-  - volumes
+    pfx: ".{{cobald_image_tag|default('latest')}}"
+  - name: volumes
 
-- copy:
+- template:
     src: cobald.Dockerfile
-    dest: /container/docker-images/cobald/Dockerfile
+    dest: "/container/docker-images/cobald.{{cobald_image_tag|default('latest')}}/Dockerfile"
     owner: "{{unpriv_user}}"
     group: docker
   register: cobald_cp_dockerfile
 
 - copy:
-    src: cobald-entrypoint.sh
+    src: "{{item}}"
-    dest: /container/docker-images/cobald/cobald-entrypoint.sh
+    dest: "/container/docker-images/cobald.{{cobald_image_tag|default('latest')}}/{{item}}"
     owner: "{{unpriv_user}}"
     group: docker
     mode: 0755
+  with_items:
+  - cobald-entrypoint.sh
+  - init-cobaldmodules.sh
   register: cobald_cp_files
 
 - docker_image:
     name: "cobald"
+    tag: "{{cobald_image_tag|default('latest')}}"
 #    pull: False
     build:
       pull: False
-      path: "/container/docker-images/cobald/"
+      path: "/container/docker-images/cobald.{{cobald_image_tag|default('latest')}}/"
     source: build
     force_source: "{{cobald_cp_dockerfile.changed or cobald_cp_files.changed}}"
 
@@ -39,7 +47,7 @@
   when: False
 
 - docker_network:
-    name: "{{docker_network}}"
+    name: "{{cobald_docker_network}}" # FIXME
     state: present
 
 # docker run -v $(pwd)/cobald-config-host:/etc/cobald -v $(pwd)/cobald:/cobald --rm -it cobald  bash
@@ -82,7 +90,7 @@
 
 - name: run pip install
   docker_container:
-    image: cobald
+    image: "cobald:{{cobald_image_tag|default('latest')}}"
     name: "cobald-src-{{item.name}}-install"
     volumes:
     - "~{{unpriv_user}}/{{item.name}}-src:/usr/local/src/{{item.name}}:rw"
@@ -102,29 +110,34 @@
 
 - import_tasks: telegraf.yml
 
-- docker_container_info:
+- name: get cobald hostname
-    name: cobald
+  include_role:
-  register: cobald_container_info
+    name: cobald_facts
+  when: cobald_container_hostname is not defined
 
-- docker_container:
+- name: run cobald container
-    name: cobald
+  docker_container:
-    image: cobald
+    name: "{{ container_name | default('cobald') }}"
-    hostname: |-
+    image: "cobald:{{cobald_image_tag|default('latest')}}"
-      {{cobald_container_info.container.Config.Hostname | default('cobald-'+
+    hostname: "{{cobald_container_hostname}}"
-        lookup('password', '/dev/null chars=ascii_lowercase length=6')) }}
+    domainname: "{{ cobald_domainname | default('cobald.local')}}"
-    domainname: cobald.local
+    volumes: "{{default_mounts + cobald_mounts }}"
-    volumes:
+    networks:
+    - name: "{{cobald_docker_network}}"
+    networks_cli_compatible: True
+#    env:
+#      slurmuser: "{{slurm_user}}"
+#    privileged: "{{ container_privileged | bool }}"
+    state: started
+    detach: True
+    cleanup: True
+    interactive: True
+    # command: python3 -m cobald.daemon /etc/cobald/config.yaml
+  vars:
+    default_mounts: "{{cobald_slurm_mounts | default([])}}"
+    cobald_mounts:
     - "~{{unpriv_user}}/cobald:/etc/cobald"
     # - /container/volumes/cobald:/etc/cobald:ro
     - "~{{unpriv_user}}/cobald/modules:/usr/local/src/cobaldmodules"
     - "~{{unpriv_user}}/cobald-src:/usr/local/src/cobald:ro"
     - "~{{unpriv_user}}/tardis-src:/usr/local/src/tardis:ro"
-    networks:
-    - name: "{{docker_network}}"
-    state: started
-    detach: True
-    cleanup: True
-    interactive: True
-    # command: /bin/bash
-    # python3 -m cobald.daemon /etc/cobald/config.yaml
-    command: python3 -m cobald.daemon /etc/cobald/config.yaml

roles/cobald/templates/cobald.Dockerfile

@@ -1,4 +1,4 @@
-FROM docker.io/library/centos:7
+FROM {{ cobald_docker_base_image | default("docker.io/library/centos:7") }}
 
 RUN yum update -y && \
 	yum install -y python3 git && pip3 install --upgrade pip && \
@@ -41,11 +41,20 @@ VOLUME /usr/local/src/cobaldmodules
 
 VOLUME /etc/cobald
 
-COPY cobald-entrypoint.sh /usr/local/sbin/docker-entrypoint.sh
+RUN mkdir -p /usr/local/lib/entrypoints.d/
 
-RUN chmod 755 /usr/local/sbin/docker-entrypoint.sh
+COPY init-cobaldmodules.sh /usr/local/lib/entrypoints.d/50-init-cobaldmodules.sh
 
-ENTRYPOINT [ "/usr/local/sbin/docker-entrypoint.sh" ]
+RUN chmod 755 /usr/local/lib/entrypoints.d/50-init-cobaldmodules.sh
+
+RUN echo -e "#!/bin/sh\npython3 -m cobald.daemon /etc/cobald/config.yaml" >> /etc/docker-init.d/70-cobald && chmod 755 /etc/docker-init.d/70-cobald
+
+{% if cobald_docker_default_command | default(True) -%}
+COPY cobald-entrypoint.sh /usr/local/sbin/cobald-entrypoint.sh
+
+RUN chmod 755 /usr/local/sbin/cobald-entrypoint.sh
+
+ENTRYPOINT [ "/usr/local/sbin/cobald-entrypoint.sh" ]
 
 RUN yum -y install iproute &&\
     yum clean all && rm -rf /var/cache/yum
@@ -54,4 +63,6 @@ USER cobald
 
 STOPSIGNAL SIGINT
 
-CMD "python3 -m cobald.daemon /etc/cobald/config.yaml"
+# CMD "python3 -m cobald.daemon /etc/cobald/config.yaml"
+CMD /etc/docker-init.d/60-cobald
+{%- endif %}

roles/cobald/vars/cobald-slurm.yml

@@ -0,0 +1,12 @@
+cobald_image_tag: slurm
+cobald_docker_base_image: "{{slurm.base_image}}"
+cobald_docker_default_command: False
+cobald_docker_network: "{{slurm.network}}"
+cobald_domainname: "{{slurm.domain}}"
+cobald_slurm_mounts: "{{slurm.mounts}}"
+#- /container/volumes/slurm/:/etc/slurm/:rw
+##- "{{slurm_cfg_path | mandatory}}:/etc/slurm/:rw"
+#- /container/volumes/munge/munge.key:/etc/munge/munge.key:rw
+## - "{{slurm_munge_path | mandatory}}:/etc/munge/munge.key:rw"
+#- slurm-shared:/shared/:rw
+## - "{{slurm_shared_path | mandatory}}:{{slurm_shared_target | default('/shared')}}:rw"

roles/cobald/vars/main.yml

@@ -0,0 +1 @@
+cobald_docker_network: "{{docker_network}}"

roles/cobald_facts/tasks/main.yml

@@ -0,0 +1,11 @@
+- block:
+  - docker_container_info:
+      name: "{{ container_name | mandatory }}"
+    register: cobald_container_info
+
+  - set_fact:
+      cobald_container_hostname: |-
+        {{cobald_container_info.container.Config.Hostname | default('cobald-'+
+          lookup('password', '/dev/null chars=ascii_lowercase length=6')) }}
+  when: cobald_container_hostname is not defined
+

roles/slurm/defaults/main.yml

@@ -3,3 +3,6 @@ slurm_user: slurm
 slurm_log_path_ctld: /var/log/slurm/slurmctld.log
 slurm_log_path_d: /var/log/slurm/slurmd.log
 slurm_log_path_sched: /var/log/slurm/slurmsched.log
+slurm_prefix: slurm
+slurm_domain: slurm.local
+docker_network: slurm

roles/slurm/files/docker-init

@@ -0,0 +1,18 @@
+#!/bin/bash
+function trp_term(){
+        echo pkill -P $pids
+        for j in $pids ; do
+                pkill -P $j
+                kill -SIGTERM $j
+        done
+}
+trap trp_term SIGINT SIGTERM
+pids=""
+for i in /etc/docker-init.d/* ; do
+        [ ! -f $i ] && break
+        $i &
+        pids="$pids $!"
+done
+wait $pids
+
+# TODO: call start scripts like "foo.sh start" and "foo.sh stop" to avoid pkill

roles/slurm/files/slurmctld.Dockerfile

@@ -1,32 +1,11 @@
-FROM docker.io/library/centos:7 as base
+FROM slurm:base
-
-RUN yum install -y epel-release && \
-	yum install -y slurm && \
-	yum clean all && rm -rf /var/cache/yum
-
-RUN	yum install -y less iproute bind-utils nmap-ncat net-tools && \
-	yum clean all && rm -rf /var/cache/yum
-
-COPY entrypoint.sh /usr/local/sbin/entrypoint.sh
-
-RUN chown root:root /usr/local/sbin/entrypoint.sh && \
-	chmod 755 /usr/local/sbin/entrypoint.sh
-
-ENTRYPOINT [ "/usr/local/sbin/entrypoint.sh" ]
-
-ARG slurmuser=slurm
-ENV slurmuser=${slurmuser}
-
-RUN useradd -d /var/lib/slurm -m --no-log-init --system $slurmuser &&\
-	slurm-setuser -u $slurmuser -g $slurmuser -y
 
 RUN	yum install -y slurm-slurmctld && \
 	yum clean all && rm -rf /var/cache/yum
 
+COPY start-scripts/20-slurmctld /etc/docker-init.d/20-slurmctld
+RUN chmod 755 /etc/docker-init.d/20-slurmctld
+
 ENV SLURMCTLD_LOG_PATH="/var/log/slurm/slurmctld.log"
 ENV SLURMD_LOG_PATH="/var/log/slurm/slurmd.log"
 ENV SLURM_SCHED_LOG_PATH="/var/log/slurm/slurmsched.log"
-
-CMD bash -c 'cat <({ su -s /bin/sh -c "munged -F" munge & \
-	su -s /bin/sh -c "slurmctld -D" ${slurmuser} 2>/dev/null 1>/dev/null & \
-	tail --retry --pid $! -f ${SLURMCTLD_LOG_PATH} ${SLURM_SCHED_LOG_PATH} & })'

roles/slurm/files/slurmd.Dockerfile

@@ -1,32 +1,11 @@
-FROM docker.io/library/centos:7
+FROM slurm:base
-
-RUN yum install -y epel-release && \
-	yum install -y slurm && \
-	yum clean all && rm -rf /var/cache/yum
-
-RUN	yum install -y less iproute bind-utils nmap-ncat net-tools && \
-	yum clean all && rm -rf /var/cache/yum
-
-COPY entrypoint.sh /usr/local/sbin/entrypoint.sh
-
-RUN chown root:root /usr/local/sbin/entrypoint.sh && \
-	chmod 755 /usr/local/sbin/entrypoint.sh
-
-ENTRYPOINT [ "/usr/local/sbin/entrypoint.sh" ]
-
-ARG slurmuser=slurm
-ENV slurmuser=${slurmuser}
-
-RUN useradd -d /var/lib/slurm -m --no-log-init --system $slurmuser &&\
-	slurm-setuser -u $slurmuser -g $slurmuser -y
 
 RUN yum install -y slurm-slurmd && \
 	yum clean all && rm -rf /var/cache/yum
 
+COPY start-scripts/30-slurmd /etc/docker-init.d/30-slurmd
+RUN chmod 755 /etc/docker-init.d/30-slurmd
+
 ENV SLURMCTLD_LOG_PATH="/var/log/slurm/slurmctld.log"
 ENV SLURMD_LOG_PATH="/var/log/slurm/slurmd.log"
 ENV SLURM_SCHED_LOG_PATH="/var/log/slurm/slurmsched.log"
-
-CMD bash -c 'cat <({ su -s /bin/sh -c "munged -F" munge & \
-	slurmd -D 2>/dev/null 1>/dev/null & \
-	tail --retry --pid $! -f ${SLURMD_LOG_PATH} ${SLURM_SCHED_LOG_PATH} & })'

roles/slurm/files/start-scripts/20-slurmctld

@@ -0,0 +1,4 @@
+#!/bin/sh
+su -s /bin/sh -c "slurmctld -D" ${slurmuser} 2>/dev/null 1>/dev/null &
+tail --retry --pid $! -f ${SLURMCTLD_LOG_PATH} ${SLURM_SCHED_LOG_PATH}
+

roles/slurm/files/start-scripts/30-slurmd

@@ -0,0 +1,4 @@
+#!/bin/sh
+slurmd -D 2>/dev/null 1>/dev/null &
+tail --retry --pid $! -f ${SLURMD_LOG_PATH} ${SLURM_SCHED_LOG_PATH}
+

roles/slurm/handlers/main.yml

@@ -1,3 +1,3 @@
 - name: reconfigure slurm 
-  command:
+  shell:
-    cmd: docker container exec -it slurm-ctl scontrol reconfigure
+    cmd: "docker container exec -it {{slurm_prefix}}-ctl scontrol reconfigure || docker container restart {{slurm_prefix}}-ctl && docker container exec -it {{slurm_prefix}}-ctl scontrol reconfigure"

roles/slurm/tasks/dockerimage.yml

@@ -1,31 +1,50 @@
 - file:
-    path: "/container/docker-images/{{item}}"
+    path: "/container/docker-images/slurm"
     state: directory
     owner: "{{unpriv_user}}"
     group: docker
 
-- copy:
+- copy: # FIXME: template
-    src: "{{item}}.Dockerfile"
+    src: "{{image.name}}.Dockerfile"
-    dest: "/container/docker-images/{{item}}/Dockerfile"
+    dest: "/container/docker-images/slurm/{{image.name}}.Dockerfile"
     owner: "{{unpriv_user}}"
     group: docker
   register: slurm_cp_dockerfile
 
-- copy:
+- name: copy entrypoint and docker-init
-    src: "entrypoint.sh"
+  copy: # FIXME: swap out
-    dest: "/container/docker-images/{{item}}/entrypoint.sh"
+    src: "{{item}}"
+    dest: "/container/docker-images/slurm/{{item}}"
     owner: root
     group: root
     mode: u=rwx,g=rx,o=rx
+  loop:
+  - entrypoint.sh
+  - docker-init
   register: slurm_cp_entrypt
 
+- name: copy startup scripts
+  copy:
+    src: "start-scripts/"
+    dest: "/container/docker-images/slurm/start-scripts/"
+    owner: root
+    group: root
+    mode: u=rwx,g=rx,o=rx
+  register: slurm_cp_stscrs
+
+- set_fact:
+    slurm_image_files_changed: "{{ (slurm_image_files_changed | default(False))
+      or slurm_cp_entrypt.changed or slurm_cp_stscrs.changed }}"
+
 - docker_image:
-    name: "slurm-{{item}}"
+    name: "slurm"
+    tag: "{{image.name}}"
 #    pull: False
     build:
       pull: False
-      path: "/container/docker-images/{{item}}"
+      path: "/container/docker-images/slurm/"
-#      target: "{{item}}" # unsupported on old docker-py versions as in el7 
+      dockerfile: "{{image.name}}.Dockerfile"
+#      target: "{{image.name}}" # unsupported on old docker-py version as in el7
     source: build
-    force_source: "{{slurm_cp_dockerfile.changed or slurm_cp_entrypt.changed}}"
+    force_source: "{{slurm_cp_dockerfile.changed or slurm_image_files_changed}}"
 

roles/slurm/tasks/main.yml

@@ -3,10 +3,22 @@
     name: [ slurm, slurm-doc ]
     state: present
 
-- include_tasks: dockerimage.yml
+- include_role:
-  loop:
+    name: slurm_dockerimage
-    - slurmctld
+  loop: # FIXME: default(omit)!
-    - slurmd
+  - name: slurmctld
+    dockerfile: "{{ lookup('file', 'slurmctld.Dockerfile') }}"
+    files:
+    - dest: start-scripts/20-slurmctld
+      content: "{{ lookup('file', 'start-scripts/20-slurmctld') }}"
+  - name: slurmd
+    dockerfile: "{{ lookup('file', 'slurmd.Dockerfile') }}"
+    files:
+    - dest: start-scripts/30-slurmd
+      content: "{{ lookup('file', 'start-scripts/30-slurmd') }}"
+  loop_control:
+    loop_var: image
+    label: "{{ image.name }}"
 
 - name: generate munge key
   shell:
@@ -47,12 +59,14 @@
   loop:
     - slurm.conf
     - cgroup.conf
+  vars:
+    alloc_nodes: "{{ [ slurm_prefix+'-submit1' ] + extra_nodes | default([])}}"
   notify: reconfigure slurm
   tags: [ slurm-config ]
 
 - name: "create docker network to make service discovery work"
   docker_network:
-    name: slurm
+    name: "{{ docker_network }}"
     state: present
 
 - name: "create docker volume for shared access between nodes"
@@ -61,13 +75,26 @@
     state: present
 
 - set_fact:
-    slurm_nodes: # default nodes: controller and submit machine
+    slurm_nodes_std: # default nodes: controller and submit machine
     - machine: ctl
-      image: slurm-slurmctld
+      image: slurm:slurmctld
+      exposed_ports: [ "6817:6817/tcp" ]
     - machine: submit1
-      image: slurm-slurmd
+      image: slurm:slurmd
       extra_mounts:
       - "/home/{{unpriv_user}}/job3/:/mnt/:rw"
+    slurm_nodes_exec: | # extend range to execute nodes list
+      {% set slurm_nodes_exec = slurm_nodes_exec | default([]) %}
+      {% for i in range(1, num_nodes+1) -%}
+        {% set _ = slurm_nodes_exec.extend([
+          {'machine':'exec%s'|format(i), 'image': 'slurm:slurmd'}]) -%}
+      {%- endfor %}
+      {{ slurm_nodes_exec }}
+    slurm_default_mounts:
+    - /container/volumes/slurm/:/etc/slurm/:rw
+    - /container/volumes/munge/munge.key:/etc/munge/munge.key:rw
+    - slurm-shared:/shared/:rw
+    slurm_network: "{{docker_network}}"
   tags: [ slurm-config ]
 
 # TODO: reserve some address using docker_network_info and assign as aux
@@ -76,30 +103,34 @@
 
 - name: run slurm docker containers
   docker_container:
-    name: "slurm-{{item.machine}}"
+    name: "{{ slurm_prefix }}-{{ item.machine }}"
-    hostname: "slurm-{{item.machine}}"
+    hostname: "{{ slurm_prefix }}-{{ item.machine }}"
-    domainname: "slurm.local"
+    domainname: "{{ slurm_domain }}"
-    volumes: "{{default_mounts + ( item.extra_mounts | default([]) ) }}"
+    volumes: "{{ slurm_default_mounts + ( item.extra_mounts | default([]) ) }}"
+    ports: "{{ item.exposed_ports | default([]) }}"
     networks:
-    - name: "slurm"
+    - name: "{{ slurm_network }}"
     env:
-      slurmuser: "{{slurm_user}}"
+      slurmuser: "{{ slurm_user }}"
-    image: "{{item.image}}"
+    image: "{{ item.image }}"
     state: started
     detach: True
     cleanup: True
     privileged: "{{ container_privileged | bool }}"
     networks_cli_compatible: True
+    interactive: True
   vars:
-    default_mounts:
+    slurm_nodes_all: "{{ slurm_nodes_exec + slurm_nodes_std }}"
-    - /container/volumes/slurm/:/etc/slurm/:rw
+  loop: "{{ slurm_nodes_all }}"
-    - /container/volumes/munge/munge.key:/etc/munge/munge.key:rw
+  loop_control:
-    - slurm-shared:/shared/:rw
+    label: "{{slurm_prefix}}-{{ item.machine }}"
-    slurm_nodes_all: | # add execute nodes
-      {% for i in range(1, 4) -%}
-        {% set _ = slurm_nodes.extend([
-          {'machine':'exec%s'|format(i), 'image': 'slurm-slurmd'}]) -%}
-      {%- endfor %}
-      {{ slurm_nodes }}
-  loop: "{{slurm_nodes_all}}"
   tags: [ slurm-config ]
+
+- name: set facts to be used by other modules
+  set_fact:
+    slurm:
+      user: "{{slurm_user}}"
+      domain: "{{slurm_domain}}"
+      base_image: "slurm:base"
+      mounts: "{{slurm_default_mounts}}"
+      network: "{{docker_network}}"

roles/slurm/templates/slurm.conf.j2

@@ -164,5 +164,7 @@ SlurmSchedLogFile={{slurm_log_path_sched}}
 #
 # COMPUTE NODES
 NodeName=slurm-exec[1-{{num_nodes}}] CPUs=2 CoresPerSocket=2 State=UNKNOWN
-NodeName=slurm-submit1 CPUs=1 State=UNKNOWN
+{% for i in alloc_nodes -%}
-PartitionName=debug Nodes=slurm-exec[1-{{num_nodes}}] AllocNodes=slurm-submit1 Default=YES MaxTime=INFINITE State=UP
+NodeName={{i}} State=UNKNOWN
+{% endfor %}
+PartitionName=debug Nodes=slurm-exec[1-{{num_nodes}}] AllocNodes={{alloc_nodes | join(',')}} Default=YES MaxTime=INFINITE State=UP

roles/slurm_dockerimage/files/entry-munge.sh

@@ -5,5 +5,3 @@ if [ -f "/etc/munge/munge.key" ] ; then
 	chown munge:munge /etc/munge/munge.key
 	chmod 600 /etc/munge/munge.key
 fi
-
-exec "$@"

roles/slurm_dockerimage/files/entrypoint.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+set -e
+
+for i in /usr/local/lib/entrypoints.d/* ; do
+    [ -f $i ] && /bin/sh $i || break
+done
+
+exec "${@:-/bin/bash}"

roles/slurm_dockerimage/files/slurm-base.Dockerfile

@@ -0,0 +1,31 @@
+FROM docker.io/library/centos:7
+
+RUN yum install -y epel-release && \
+	yum install -y slurm && \
+	yum clean all && rm -rf /var/cache/yum
+
+RUN	yum install -y less iproute bind-utils nmap-ncat net-tools && \
+	yum clean all && rm -rf /var/cache/yum
+
+RUN mkdir -p /usr/local/lib/entrypoints.d/
+
+COPY --chown=root:root entry-munge.sh /usr/local/lib/entrypoints.d/10-munge.sh
+COPY --chown=root:root entrypoint.sh /usr/local/sbin/entrypoint.sh
+
+RUN chmod 755 /usr/local/lib/entrypoints.d/10-munge.sh && \
+	chmod 755 /usr/local/sbin/entrypoint.sh
+
+ENTRYPOINT [ "/usr/local/sbin/entrypoint.sh" ]
+
+ARG slurmuser=slurm
+ENV slurmuser=${slurmuser}
+
+RUN useradd -d /var/lib/slurm -m --no-log-init --system $slurmuser &&\
+	slurm-setuser -u $slurmuser -g $slurmuser -y
+
+COPY docker-init /usr/local/sbin/docker-init
+RUN mkdir /etc/docker-init.d  && chmod 755 /usr/local/sbin/docker-init
+COPY start-scripts/10-munge /etc/docker-init.d/10-munge
+RUN chmod 755 /etc/docker-init.d/10-munge
+
+CMD /usr/local/sbin/docker-init

roles/slurm_dockerimage/files/start-scripts/10-munge

@@ -0,0 +1,2 @@
+#!/bin/sh
+exec su -s /bin/sh -c "munged -F" munge

roles/slurm_dockerimage/tasks/dockerimage.yml

@@ -0,0 +1,40 @@
+- name: create directories for docker image build
+  file:
+    path: "/container/docker-images/slurm-{{image.name}}/{{item}}"
+    state: directory
+    owner: "{{unpriv_user}}"
+    group: docker
+  loop: "{{ [''] + (image.files | map(attribute='dest') | map('dirname') |
+                    unique | select | list) }}"
+
+- name: copy Dockerfile
+  copy:
+    src: "{{image.name}}.Dockerfile"
+    dest: "/container/docker-images/slurm-{{image.name}}/Dockerfile"
+    owner: "{{unpriv_user}}"
+    group: docker
+  register: slurm_cp_dockerfile
+
+- name: copy requisite files
+  copy:
+    content: "{{ item.content }}"
+    dest: "/container/docker-images/slurm-{{image.name}}/{{item.dest}}"
+    owner: root
+    group: root
+    mode: u=rwx,g=rx,o=rx
+  loop: "{{ image.files | default([]) }}"
+  loop_control:
+    label: "{{ item.dest }}"
+  register: slurm_cp_files
+
+- docker_image:
+    name: "slurm"
+    tag: "{{image.name}}"
+#    pull: False
+    build:
+      pull: False
+      path: "/container/docker-images/slurm-{{image.name}}/"
+    source: build
+    force_source: "{{slurm_cp_dockerfile.changed or
+                     slurm_cp_files.changed or
+                     slurm_baseimg_build_chg }}"

roles/slurm_dockerimage/tasks/main.yml

@@ -0,0 +1,46 @@
+- file:
+    path: "/container/docker-images/slurm-base/start-scripts"
+    state: directory
+    owner: "{{unpriv_user}}"
+    group: docker
+
+# - name: copy Dockerfile, entrypoint, docker-init and munge startup
+- name: copy slurm base image requisite files
+  copy: # FIXME: swap out
+    src: "{{item.file}}"
+    dest: "/container/docker-images/slurm-base/{{item.file}}"
+    owner: "{{unpriv_user}}"
+    group: docker
+    mode: "{{ item.perms | default('u=rwx,g=rx,o=rx') }}"
+  loop:
+  - file: slurm-base.Dockerfile
+    perms: u=rw,g=r,o=r
+  - file: entrypoint.sh
+  - file: entry-munge.sh
+  - file: docker-init
+  - file: start-scripts/10-munge
+  when: not (slurm_baseimg_build_chg | default(False))
+  register: slurm_baseimg_copy
+
+- name: build base image
+  docker_image:
+    name: "slurm"
+    tag: "base"
+#    pull: False
+    build:
+      pull: False
+      path: "/container/docker-images/slurm-base/"
+      dockerfile: "slurm-base.Dockerfile"
+#      target: "{{image.name}}" # unsupported on old docker-py version as in el7
+    source: build
+    force_source: "{{slurm_baseimg_copy.changed}}"
+  # when: run only once but keep changed state
+  when: not (slurm_baseimg_build_chg | default(False))
+  register: slurm_baseimg_build
+
+- set_fact:
+    slurm_baseimg_build_chg:
+      "{{(slurm_baseimg_build_chg | default(False)) or
+         slurm_baseimg_build.changed}}"
+
+- include_tasks: dockerimage.yml