slurm container running when drone started

This commit is contained in:
2021-07-01 15:19:35 +02:00
parent d88761ca7d
commit 8bc2f717e0
8 changed files with 114 additions and 4 deletions

View File

@@ -0,0 +1,3 @@
#!/bin/sh
[ /slurm-singimage/slurmd.sif -nt /shared/slurmd.sif ] && \
cp /slurm-singimage/slurmd.sif /shared/slurmd.sif

View File

@@ -0,0 +1,3 @@
#!/bin/sh
slurmd --conf-server ${slurmctld} -D -N ${nodename} 2>/dev/null 1>/dev/null &
tail --retry --pid $! -f ${SLURMD_LOG_PATH} ${SLURM_SCHED_LOG_PATH}

View File

@@ -0,0 +1,30 @@
Bootstrap: docker-daemon
From: slurm:slurmd
%files
31-slurmd-configless /etc/docker-init.d/31-slurm-configless
/container/volumes/munge/munge.key /etc/munge/munge.key
%post
rm /etc/docker-init.d/30-slurmd
chmod 755 /etc/docker-init.d/31-slurm-configless
%startscript
if [ -z "${1}" -o -z "${2}" ] ; then
echo "undefined variables slurmctld or nodename"
exit 1
fi
export slurmctld="${1}"
export nodename="${2}"
echo ${slurmctld} ${nodename} ${SLURMD_LOG_PATH} ${SLURM_SCHED_LOG_PATH}
exec /usr/local/sbin/entrypoint.sh /usr/local/sbin/docker-init
%runscript
if [ -z "${1}" -o -z "${2}" ] ; then
echo "undefined variables slurmctld or nodename"
exit 1
fi
export slurmctld="${1}"
export nodename="${2}"
echo ${slurmctld} ${nodename} ${SLURMD_LOG_PATH} ${SLURM_SCHED_LOG_PATH}
exec /usr/local/sbin/entrypoint.sh /usr/local/sbin/docker-init

View File

@@ -0,0 +1,25 @@
#!/bin/sh
#SBATCH -D /shared
export
echo $@
nodename=$(hostname | awk '{ print "drone" substr($1,match($1, "([[:digit:]]+)")) }')
function handler_quit(){
echo "shutdown container"
singularity instance stop slurm-drone
scontrol update NodeName=${nodename} State=FUTURE
exit 0
}
trap handler_quit EXIT
# set -x
echo "starting ${nodename}"
singularity instance start --writable-tmpfs /shared/slurmd.sif slurm-drone \
slurm-ctl ${nodename}
if [ $? -eq 0 ] ; then
echo "container started, sleeping $(( 60 * ${SLURM_Walltime}))"
sleep $(( 60 * ${SLURM_Walltime} ))
fi
handler_quit

View File

@@ -15,6 +15,10 @@
content: "{{ lookup('file', 'cobald-entrypoint.sh') }}"
- dest: init-cobaldmodules.sh
content: "{{ lookup('file', 'init-cobaldmodules.sh') }}"
- dest: start-drone
content: "{{ lookup('file', 'start-drone') }}"
- dest: 28-sync-container-slurmd
content: "{{ lookup('file', '28-sync-container-slurmd') }}"
when: cobald_slurm | default(False)
- name: build generic cobald docker image
@@ -136,6 +140,7 @@
cobald_mounts:
- "~{{unpriv_user}}/cobald:/etc/cobald"
# - /container/volumes/cobald:/etc/cobald:ro
- "/container/docker-images/sing-slurmd/build/:/slurm-singimage/: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"

View File

@@ -1,8 +1,47 @@
- name: "setup singularity"
- name: setup singularity
import_role: name="singularity"
tags: singularity
#- name: copy slurm def file
# - name: export docker image
#- name: build container
- name: make singularity image build directory
file:
state: directory
path: "{{item}}"
owner: "{{unpriv_user}}"
group: "docker"
mode: "0755"
loop:
- /container/docker-images/sing-slurmd
- /container/docker-images/sing-slurmd/cache
- /container/docker-images/sing-slurmd/build
- name: copy slurm singularity container files
copy:
src: "{{item}}"
dest: "/container/docker-images/sing-slurmd/{{item}}"
owner: "{{unpriv_user}}"
group: "docker"
loop:
- slurm-slurmd.def
- 31-slurmd-configless
register: cobald_copy_sing_files
- name: remove old container
file:
path: /container/docker-images/sing-slurmd/build/slurmd.sif
state: absent
when: cobald_copy_sing_files.changed
- name: build container
shell:
chdir: /container/docker-images/sing-slurmd/
cmd: SINGULARITY_TMPDIR=/container/docker-images/sing-slurmd/cache
singularity build --disable-cache
/container/docker-images/sing-slurmd/build/slurmd.sif
/container/docker-images/sing-slurmd/slurm-slurmd.def
creates: /container/docker-images/sing-slurmd/build/slurmd.sif
register: cobald_sing_build
- debug: msg="{{[cobald_sing_build.stdout, cobald_sing_build.stderr]}}"
tags: [ never, debug ]
# TODO: trigger copy in cobald container when slurmd.sif rebuilt

View File

@@ -48,6 +48,10 @@ COPY init-cobaldmodules.sh /usr/local/lib/entrypoints.d/50-init-cobaldmodules.sh
RUN chmod 755 /usr/local/lib/entrypoints.d/50-init-cobaldmodules.sh
COPY start-drone /usr/local/bin/start-drone
COPY 28-sync-container-slurmd /etc/docker-init.d/28-sync-container-slurmd
RUN chmod 755 /usr/local/bin/start-drone /etc/docker-init.d/28-sync-container-slurmd
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) -%}