Created
July 12, 2022 11:07
-
-
Save adrienbernede/7e0f57042fdb65d220061a8ef57b0c24 to your computer and use it in GitHub Desktop.
A minimal example of shared allocation between CI jobs on Gitlab (using Slurm)
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
############################################################################## | |
# Copyright (c) 2022, Lawrence Livermore National Security, LLC and RADIUSS | |
# project contributors. See the COPYRIGHT file for details. | |
# | |
# SPDX-License-Identifier: (MIT) | |
############################################################################## | |
# The allocation name must be unique | |
variables: | |
ALLOC_NAME: ${CI_PROJECT_NAME}_ci_${CI_PIPELINE_ID} | |
stages: | |
- allocate | |
- run | |
- release | |
.on_ruby: | |
tags: | |
- shell | |
- ruby | |
allocate_resources (on ruby): | |
extends: .on_ruby | |
variables: | |
GIT_STRATEGY: none | |
stage: allocate | |
script: | |
- salloc --partition=pdebug --time=10 --nodes=1 --cpus-per-task=36 --no-shell --job-name=${ALLOC_NAME} | |
run:1: | |
extends: .on_ruby | |
# With slum versions >= 20.11 it is mandatory to specify whether we want concurrent/overlapping jobs | |
variables: | |
SLURM_OVERLAP: 1 | |
stage: run | |
script: | |
# Retrieving the allocation name in a separate shell is the uglier part | |
- export JOBID=$(squeue -h --name=${ALLOC_NAME} --format=%A) | |
# Allocation options here serve for scheduling inside the main allocation | |
- srun $( [[ -n "${JOBID}" ]] && echo "--jobid=${JOBID}" ) --time=1 --nodes=1 ./script | |
run:2: | |
extends: .on_ruby | |
variables: | |
SLURM_OVERLAP: 1 | |
stage: run | |
script: | |
- export JOBID=$(squeue -h --name=${ALLOC_NAME} --format=%A) | |
- srun $( [[ -n "${JOBID}" ]] && echo "--jobid=${JOBID}" ) --time=1 --nodes=1 ./script | |
run:3: | |
extends: .on_ruby | |
variables: | |
SLURM_OVERLAP: 1 | |
stage: run | |
script: | |
- export JOBID=$(squeue -h --name=${ALLOC_NAME} --format=%A) | |
- srun $( [[ -n "${JOBID}" ]] && echo "--jobid=${JOBID}" ) --time=1 --nodes=1 ./script | |
release_resources (on ruby): | |
variables: | |
GIT_STRATEGY: none | |
extends: .on_ruby | |
stage: release | |
script: | |
- export JOBID=$(squeue -h --name=${ALLOC_NAME} --format=%A) | |
- ([[ -n "${JOBID}" ]] && scancel ${JOBID}) | |
# Releasing the allocation must be done even of pipeline failure. | |
when: always | |
#### script example for demo: #### | |
# | |
# #!/bin/bash | |
# | |
# echo "## START: $(date)" | |
# echo "This script is running on $(hostname)" | |
# echo "Let’s sleep for 3 seconds..." | |
# sleep 3 | |
# echo "## END: $(date)" |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment