- Notifications
You must be signed in to change notification settings - Fork47
Slurm in Docker - Exploring Slurm using CentOS 7 based Docker images
License
NotificationsYou must be signed in to change notification settings
SciDAS/slurm-in-docker
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
WORK IN PROGRESS
UseDocker to explore the various components ofSlurm
This work represents a small exploratory Slurm cluster using CentOS 7 based Docker images. The intent was to learn the basics of Slurm prior to extending the concept to a more distributed environment.
Images include:
- Slurm 19.05.1 - installed fromrpm packages
- OpenMPI 3.0.1 - installed fromrpm packages
- Lmod 7.7 - installed fromdistribution files
- Lmod module packages for CentOS 7 - Organized for Slurm-in-Docker use
- Using Lmod with Slurm-in-Docker documentation
- packages - Build the RPM packages for running Slurm and OpenMPI on CentOS 7
- base - Slurm base image from which other components are derived
- controller - Slurm controller (head-node) definition
- database - Slurm database definition (not necessary, but useful for accounting information)
- worker - Slurm worker (compute-node) definition
An exampledocker-compose.yml file is provided that builds and deploys the diagramed topology
Listing of participating containers with FQDNs and their function within the cluster.
| Container | Function | FQDN |
|---|---|---|
| controller | Slurm Primary Controller | controller.local.dev |
| database | Slurm Primary Database Daemon | database.local.dev |
| worker01 | Slurm Worker | worker01.local.dev |
| worker02 | Slurm Worker | worker02.local.dev |
Users may use the default slurm.conf file generated indocker-entrypoint.sh, or preferably create one to better fit their system.
TheSlurm Configuration Tool is a useful resource for creating custom slurm.conf files.
Steps to add user profided slurm.conf/slurmdbd.conf:
- Create
home/configandsecretdirectories:
mkdir -p home/config secret- Copy configuration files to the
home/configdirectory:
cp <user-provided-slurm.conf> home/config/slurm.conf; cp <user-provided-slurmdbd.conf> home/config/slurmdbd.confThe user can then proceed as normal.
TODO: Have software check validity of custom configuration files.
Build the slurm RPM files by following the instructions in thepackages directory.
Create the base Slurm image:
Copy thepackages/centos-7/rpms directory to thebase directory
cd base/cp -r ../packages/centos-7/rpms .Build the base image
docker build -t scidas/slurm.base:19.05.1 .Verify image build
$docker imagesREPOSITORY TAG IMAGE ID CREATED SIZEscidas/slurm.base 19.05.1 1600621cb483 Less than a second ago 819MB...
All images defined indocker-compose.yml will be built from thescidas/slurm.base:19.05.1 base image
An exampledocker-compose.yml file is provided that builds and deploys the diagramed topology (-d is used to daemonize the call).
docker-compose up -dFour containers should be observed running when completed
$docker psCONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES995183e9391e scidas/slurm.worker:19.05.1 "/usr/local/bin/tini…" 10 seconds ago Up 30 seconds 22/tcp, 3306/tcp, 6817-6819/tcp, 60001-63000/tcp worker01bdd7c8daaca2 scidas/slurm.database:19.05.1 "/usr/local/bin/tini…" 10 seconds ago Up 30 seconds 22/tcp, 3306/tcp, 6817-6819/tcp, 60001-63000/tcp databasea8382a486989 scidas/slurm.worker:19.05.1 "/usr/local/bin/tini…" 10 seconds ago Up 30 seconds 22/tcp, 3306/tcp, 6817-6819/tcp, 60001-63000/tcp worker0224e951854109 scidas/slurm.controller:19.05.1 "/usr/local/bin/tini…" 11 seconds ago Up 31 seconds 22/tcp, 3306/tcp, 6817-6819/tcp, 60001-63000/tcp controller
The examples make use of the following commands.
Use thedocker exec call to gain a shell on thecontroller container.
$dockerexec -ti controller /bin/bash[root@controller /]#
Issue ansinfo call
#sinfo -lNWed Apr 11 21:15:35 2018NODELIST NODES PARTITION STATE CPUS S:C:T MEMORY TMP_DISK WEIGHT AVAIL_FE REASONworker01 1 docker* idle 1 1:1:1 1998 0 1 (null) noneworker02 1 docker* idle 1 1:1:1 1998 0 1 (null) none
Create aworker account andworker user in Slurm
#sacctmgr -i add account worker description="worker account" Organization=Slurm-in-Docker Adding Account(s) worker Settings Description = worker account Organization = slurm-in-docker Associations A = worker C = snowflake Settings Parent = root#sacctmgr -i create user worker account=worker adminlevel=None Adding User(s) worker Settings = Admin Level = None Associations = U = worker A = worker C = snowflake Non Default Settings
Use thedocker exec call to gain a MariaDB/MySQL shell on thedatabase container.
$dockerexec -ti database mysql -uslurm -ppassword -hdatabase.local.devWelcome to the MariaDB monitor. Commands end with ; or \g.Your MariaDB connection id is 9Server version: 5.5.56-MariaDB MariaDB ServerCopyright (c) 2000, 2017, Oracle, MariaDB Corporation Ab and others.Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.MariaDB [(none)]>
Checkout theslurm_acct_db database and it's tables
MariaDB [(none)]> use slurm_acct_db;Reading table information for completion of table and column namesYou can turn off this feature to get a quicker startup with -ADatabase changedMariaDB [slurm_acct_db]> show tables;+-----------------------------------+| Tables_in_slurm_acct_db |+-----------------------------------+| acct_coord_table || acct_table || clus_res_table || cluster_table || convert_version_table || federation_table || qos_table || res_table || snowflake_assoc_table || snowflake_assoc_usage_day_table || snowflake_assoc_usage_hour_table || snowflake_assoc_usage_month_table || snowflake_event_table || snowflake_job_table || snowflake_last_ran_table || snowflake_resv_table || snowflake_step_table || snowflake_suspend_table || snowflake_usage_day_table || snowflake_usage_hour_table || snowflake_usage_month_table || snowflake_wckey_table || snowflake_wckey_usage_day_table || snowflake_wckey_usage_hour_table || snowflake_wckey_usage_month_table || table_defs_table || tres_table || txn_table || user_table |+-----------------------------------+29 rows in set (0.00 sec)
Validate that theworker user was entered into the database
MariaDB [slurm_acct_db]> select * from user_table;+---------------+------------+---------+--------+-------------+| creation_time | mod_time | deleted | name | admin_level |+---------------+------------+---------+--------+-------------+| 1523481120 | 1523481120 | 0 | root | 3 || 1523481795 | 1523481795 | 0 | worker | 1 |+---------------+------------+---------+--------+-------------+2 rows in set (0.00 sec)
Use thedocker exec call to gain a shell on either theworker01 orworker02 container and become the userworker.
$dockerexec -ti -u worker worker01 /bin/bash[worker@worker01 /]$cd~[worker@worker01 ~]$pwd/home/worker
Test password-lessssh between containers
[worker@worker01 ~]$hostnameworker01.local.dev[worker@worker01 ~]$ssh worker02[worker@worker02 ~]$hostnameworker02.local.dev[worker@worker02 ~]$ssh controller[worker@controller ~]$hostnamecontroller.local.dev
All commands are issued as the userworker from thecontroller node
$dockerexec -ti -u worker controller /bin/bash[worker@controller /]$cd~[worker@controller ~]$pwd/home/worker
- For the rest of this section the
[worker@controller ~]$prompt will be shortend to simply$
Test thesacct andsrun calls
$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------$srun -N 2 hostnameworker01.local.devworker02.local.dev$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 hostname docker worker 2 COMPLETED 0:0
Test thesbatch call
Make a job file named:slurm_test.job
#!/bin/bash#SBATCH --job-name=SLURM_TEST#SBATCH --output=SLURM_TEST.out#SBATCH --error=SLURM_TEST.err#SBATCH --partition=dockersrun hostname| sort
Run the job usingsbatch
$sbatch -N 2 slurm_test.jobSubmitted batch job 3
Check thesacct output
$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 hostname docker worker 2 COMPLETED 0:03 SLURM_TEST docker worker 2 COMPLETED 0:03.batch batch worker 1 COMPLETED 0:03.0 hostname worker 2 COMPLETED 0:0
Check the output files
$ls -1SLURM_TEST.errSLURM_TEST.outslurm_test.job$cat SLURM_TEST.outworker01.local.devworker02.local.dev
Test thesbatch --array andsqueue calls
Make a job file namedarray_test.job:
#!/bin/bash#SBATCH -N 1#SBATCH -c 1#SBATCH -t 24:00:00##################### %A == SLURM_ARRAY_JOB_ID## %a == SLURM_ARRAY_TASK_ID (or index)## %N == SLURMD_NODENAME (directories made ahead of time)#SBATCH -o %N/%A_%a_out.txt#SBATCH -e %N/%A_%a_err.txtsnooze=$(( ( RANDOM%10 )+1))echo"$(hostname) is snoozing for${snooze} seconds..."sleep$snooze
This job defines output directories as being%N which reflect theSLURMD_NODENAME variable. The output directories will need to exist ahead of time in this particular case, and can be determined by finding all available nodes in theNODELIST and creating the directories.
$sinfo -NNODELIST NODES PARTITION STATEworker01 1 docker* idleworker02 1 docker* idle$mkdir worker01 worker02
The job when run will direct it's output files to the directory defined by the node on which it is running. Each iteration will sleep from 1 to 10 seconds randomly before moving onto the next run in the array.
We will run an array of 20 jobs, 2 at a time, until the array is completed. The status can be found using thesqueue command.
$sbatch --array=1-20%2 array_test.jobSubmitted batch job 4$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 4_[3-20%2] docker array_te worker PD 0:00 1 (JobArrayTaskLimit) 4_1 docker array_te worker R 0:01 1 worker01 4_2 docker array_te worker R 0:01 1 worker02...$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 4_[20%2] docker array_te worker PD 0:00 1 (JobArrayTaskLimit) 4_19 docker array_te worker R 0:04 1 worker02 4_18 docker array_te worker R 0:10 1 worker01$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
Looking into each of theworker01 andworker02 directories we can see which jobs were run on each node.
$lsSLURM_TEST.err array_test.job worker01SLURM_TEST.out slurm_test.job worker02$ls worker014_11_err.txt 4_16_err.txt 4_1_err.txt 4_3_err.txt 4_7_err.txt4_11_out.txt 4_16_out.txt 4_1_out.txt 4_3_out.txt 4_7_out.txt4_14_err.txt 4_18_err.txt 4_20_err.txt 4_5_err.txt 4_9_err.txt4_14_out.txt 4_18_out.txt 4_20_out.txt 4_5_out.txt 4_9_out.txt$ls worker024_10_err.txt 4_13_err.txt 4_17_err.txt 4_2_err.txt 4_6_err.txt4_10_out.txt 4_13_out.txt 4_17_out.txt 4_2_out.txt 4_6_out.txt4_12_err.txt 4_15_err.txt 4_19_err.txt 4_4_err.txt 4_8_err.txt4_12_out.txt 4_15_out.txt 4_19_out.txt 4_4_out.txt 4_8_out.txt
And looking at each*_out.txt file view the output
$cat worker01/4_14_out.txtworker01.local.dev is snoozing for 10 seconds...$cat worker02/4_6_out.txtworker02.local.dev is snoozing for 7 seconds...
Using thesacct call we can see when each job in the array was executed
$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 hostname docker worker 2 COMPLETED 0:03 SLURM_TEST docker worker 2 COMPLETED 0:03.batch batch worker 1 COMPLETED 0:03.0 hostname worker 2 COMPLETED 0:04_20 array_tes+ docker worker 1 COMPLETED 0:04_20.batch batch worker 1 COMPLETED 0:04_1 array_tes+ docker worker 1 COMPLETED 0:04_1.batch batch worker 1 COMPLETED 0:04_2 array_tes+ docker worker 1 COMPLETED 0:04_2.batch batch worker 1 COMPLETED 0:04_3 array_tes+ docker worker 1 COMPLETED 0:04_3.batch batch worker 1 COMPLETED 0:04_4 array_tes+ docker worker 1 COMPLETED 0:04_4.batch batch worker 1 COMPLETED 0:04_5 array_tes+ docker worker 1 COMPLETED 0:04_5.batch batch worker 1 COMPLETED 0:04_6 array_tes+ docker worker 1 COMPLETED 0:04_6.batch batch worker 1 COMPLETED 0:04_7 array_tes+ docker worker 1 COMPLETED 0:04_7.batch batch worker 1 COMPLETED 0:04_8 array_tes+ docker worker 1 COMPLETED 0:04_8.batch batch worker 1 COMPLETED 0:04_9 array_tes+ docker worker 1 COMPLETED 0:04_9.batch batch worker 1 COMPLETED 0:04_10 array_tes+ docker worker 1 COMPLETED 0:04_10.batch batch worker 1 COMPLETED 0:04_11 array_tes+ docker worker 1 COMPLETED 0:04_11.batch batch worker 1 COMPLETED 0:04_12 array_tes+ docker worker 1 COMPLETED 0:04_12.batch batch worker 1 COMPLETED 0:04_13 array_tes+ docker worker 1 COMPLETED 0:04_13.batch batch worker 1 COMPLETED 0:04_14 array_tes+ docker worker 1 COMPLETED 0:04_14.batch batch worker 1 COMPLETED 0:04_15 array_tes+ docker worker 1 COMPLETED 0:04_15.batch batch worker 1 COMPLETED 0:04_16 array_tes+ docker worker 1 COMPLETED 0:04_16.batch batch worker 1 COMPLETED 0:04_17 array_tes+ docker worker 1 COMPLETED 0:04_17.batch batch worker 1 COMPLETED 0:04_18 array_tes+ docker worker 1 COMPLETED 0:04_18.batch batch worker 1 COMPLETED 0:04_19 array_tes+ docker worker 1 COMPLETED 0:04_19.batch batch worker 1 COMPLETED 0:0
The examples make use of the following commands.
All commands are issued as the userworker from thecontroller node
$dockerexec -ti -u worker controller /bin/bash[worker@controller /]$cd~[worker@controller ~]$pwd/home/worker
Available implementions of MPI
$srun --mpi=listsrun: MPI types are...srun: nonesrun: pmi2srun: openmpi
About Open MPI
$ompi_info Package: Open MPI root@a6fd2549e449 Distribution Open MPI: 3.0.1 Open MPI repo revision: v3.0.1 Open MPI release date: Mar 29, 2018 Open RTE: 3.0.1 Open RTE repo revision: v3.0.1 Open RTE release date: Mar 29, 2018 OPAL: 3.0.1 OPAL repo revision: v3.0.1 OPAL release date: Mar 29, 2018 MPI API: 3.1.0 Ident string: 3.0.1 Prefix: /usr Configured architecture: x86_64-redhat-linux-gnu Configure host: a6fd2549e449 Configured by: root Configured on: Fri Apr 13 02:32:11 UTC 2018 Configure host: a6fd2549e449 Configure command line: '--build=x86_64-redhat-linux-gnu' '--host=x86_64-redhat-linux-gnu' '--program-prefix=' '--disable-dependency-tracking' '--prefix=/usr' '--exec-prefix=/usr' '--bindir=/usr/bin' '--sbindir=/usr/sbin' '--sysconfdir=/etc' '--datadir=/usr/share' '--includedir=/usr/include' '--libdir=/usr/lib64' '--libexecdir=/usr/libexec' '--localstatedir=/var' '--sharedstatedir=/var/lib' '--mandir=/usr/share/man' '--infodir=/usr/share/info' '--with-slurm' '--with-pmi' '--with-libfabric=' 'LDFLAGS=-Wl,--build-id -Wl,-rpath -Wl,/lib64 -Wl,--enable-new-dtags'...
Hello world usingmpi_hello.out
Create a new file calledmpi_hello.c in/home/worker and compile it:
/****************************************************************************** * * FILE: mpi_hello.c * * DESCRIPTION: * * MPI tutorial example code: Simple hello world program * * AUTHOR: Blaise Barney * * LAST REVISED: 03/05/10 * ******************************************************************************/#include<mpi.h>#include<stdio.h>#include<stdlib.h>#defineMASTER 0intmain (intargc,char*argv[]) {intnumtasks,taskid,len;charhostname[MPI_MAX_PROCESSOR_NAME];MPI_Init(&argc,&argv);MPI_Comm_size(MPI_COMM_WORLD,&numtasks);MPI_Comm_rank(MPI_COMM_WORLD,&taskid);MPI_Get_processor_name(hostname,&len);printf ("Hello from task %d on %s!\n",taskid,hostname);if (taskid==MASTER)printf("MASTER: Number of MPI tasks is: %d\n",numtasks);//while(1) {}MPI_Finalize();}
$mpicc mpi_hello.c -o mpi_hello.out$ls| grep mpimpi_hello.cmpi_hello.out
Testmpi_hello.out using the MPI versions avalaible on the system withsrun
single node usingopenmpi
$srun --mpi=openmpi mpi_hello.outHello from task 0 on worker01.local.dev!MASTER: Number of MPI tasks is: 1$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 mpi_hello+ docker worker 1 COMPLETED 0:0
two nodes usingopenmpi
$srun -N 2 --mpi=openmpi mpi_hello.outHello from task 0 on worker01.local.dev!MASTER: Number of MPI tasks is: 2Hello from task 1 on worker02.local.dev!$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 mpi_hello+ docker worker 1 COMPLETED 0:03 mpi_hello+ docker worker 2 COMPLETED 0:0
two nodes usingpmi2
$srun -N 2 --mpi=pmi2 mpi_hello.outHello from task 0 on worker01.local.dev!MASTER: Number of MPI tasks is: 2Hello from task 1 on worker02.local.dev!$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------2 mpi_hello+ docker worker 1 COMPLETED 0:03 mpi_hello+ docker worker 2 COMPLETED 0:04 mpi_hello+ docker worker 2 COMPLETED 0:0
Run a batch array with a sleep to observe the queue
Create a file namedmpi_batch.job in/home/worker (similar to the script used for thesbatch --array example from above, and make an output directory namedmpi_out)
filempi_batch.job:
#!/bin/bash#SBATCH -N 1#SBATCH -c 1#SBATCH -t 24:00:00##################### %A == SLURM_ARRAY_JOB_ID## %a == SLURM_ARRAY_TASK_ID (or index)#SBATCH -o mpi_out/%A_%a_out.txt#SBATCH -e mpi_out/%A_%a_err.txtsnooze=$(( ( RANDOM%10 )+1))sleep$snoozesrun -N 2 --mpi=openmpi mpi_hello.out
Make directorympi_out
$mkdir mpi_outRun ansbatch array of 5 jobs, one at a time, using both nodes.
$sbatch -N 2 --array=1-5%1 mpi_batch.jobSubmitted batch job 10$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 10_[2-5%1] docker mpi_batc worker PD 0:00 2 (JobArrayTaskLimit) 10_1 docker mpi_batc worker R 0:03 2 worker[01-02]$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------...10_[2-5%1] mpi_batch+ docker worker 2 PENDING 0:010_1 mpi_batch+ docker worker 2 COMPLETED 0:010_1.batch batch worker 1 COMPLETED 0:010_1.0 mpi_hello+ worker 2 COMPLETED 0:0
...
$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 10_[4-5%1] docker mpi_batc worker PD 0:00 2 (JobArrayTaskLimit) 10_3 docker mpi_batc worker R 0:05 2 worker[01-02]$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------...10_[4-5%1] mpi_batch+ docker worker 2 PENDING 0:010_1 mpi_batch+ docker worker 2 COMPLETED 0:010_1.batch batch worker 1 COMPLETED 0:010_1.0 mpi_hello+ worker 2 COMPLETED 0:010_2 mpi_batch+ docker worker 2 COMPLETED 0:010_2.batch batch worker 1 COMPLETED 0:010_2.0 mpi_hello+ worker 2 COMPLETED 0:010_3 mpi_batch+ docker worker 2 COMPLETED 0:010_3.batch batch worker 1 COMPLETED 0:010_3.0 mpi_hello+ worker 2 COMPLETED 0:0
...
$squeue JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)$sacct JobID JobName Partition Account AllocCPUS State ExitCode------------ ---------- ---------- ---------- ---------- ---------- --------...10_5 mpi_batch+ docker worker 2 COMPLETED 0:010_5.batch batch worker 1 COMPLETED 0:010_5.0 mpi_hello+ worker 2 COMPLETED 0:010_1 mpi_batch+ docker worker 2 COMPLETED 0:010_1.batch batch worker 1 COMPLETED 0:010_1.0 mpi_hello+ worker 2 COMPLETED 0:010_2 mpi_batch+ docker worker 2 COMPLETED 0:010_2.batch batch worker 1 COMPLETED 0:010_2.0 mpi_hello+ worker 2 COMPLETED 0:010_3 mpi_batch+ docker worker 2 COMPLETED 0:010_3.batch batch worker 1 COMPLETED 0:010_3.0 mpi_hello+ worker 2 COMPLETED 0:010_4 mpi_batch+ docker worker 2 COMPLETED 0:010_4.batch batch worker 1 COMPLETED 0:010_4.0 mpi_hello+ worker 2 COMPLETED 0:0
Check thempi_out output directory
$ls mpi_out/10_1_err.txt 10_2_err.txt 10_3_err.txt 10_4_err.txt 10_5_err.txt10_1_out.txt 10_2_out.txt 10_3_out.txt 10_4_out.txt 10_5_out.txt$cat mpi_out/10_3_out.txtHello from task 1 on worker02.local.dev!Hello from task 0 on worker01.local.dev!MASTER: Number of MPI tasks is: 2
The containers, networks, and volumes associated with the cluster can be torn down by simply running:
./teardown.shEach step of this teardown may also be run individually:
The containers can be stopped and removed usingdocker-compose
$docker-compose stopStopping worker01 ... doneStopping database ... doneStopping worker02 ... doneStopping controller ... done$docker-compose rm -fGoing to remove worker01, database, worker02, controllerRemoving worker01 ... doneRemoving database ... doneRemoving worker02 ... doneRemoving controller ... done
The network and volumes can be removed using their representativedocker commands
Volumes
$docker volume listDRIVER VOLUME NAME...local slurmindocker_homelocal slurmindocker_secret$docker volume rm slurmindocker_home slurmindocker_secretslurmindocker_homeslurmindocker_secret
Network
$docker network listNETWORK ID NAME DRIVER SCOPE...a94c168fb653 slurmindocker_slurm bridge local$docker network rm slurmindocker_slurmslurmindocker_slurm
Slurm workload manager:https://www.schedmd.com/index.php
- Slurm is a highly configurable open-source workload manager. In its simplest configuration, it can be installed and configured in a few minutes (seeCaos NSA and Perceus: All-in-one Cluster Software Stack by Jeffrey B. Layton). Use of optional plugins provides the functionality needed to satisfy the needs of demanding HPC centers. More complex configurations rely upon a database for archiving accounting records, managing resource limits by user or bank account, and supporting sophisticated scheduling algorithms.
Docker:https://www.docker.com
- Docker is the company driving the container movement and the only container platform provider to address every application across the hybrid cloud. Today’s businesses are under pressure to digitally transform but are constrained by existing applications and infrastructure while rationalizing an increasingly diverse portfolio of clouds, datacenters and application architectures. Docker enables true independence between applications and infrastructure and developers and IT ops to unlock their potential and creates a model for better collaboration and innovation.
OpenMPI:https://www.open-mpi.org
- The Open MPI Project is an open sourceMessage Passing Interface implementation that is developed and maintained by a consortium of academic, research, and industry partners. Open MPI is therefore able to combine the expertise, technologies, and resources from all across the High Performance Computing community in order to build the best MPI library available. Open MPI offers advantages for system and software vendors, application developers and computer science researchers.
Lmod:http://lmod.readthedocs.io/en/latest/index.html
- Lmod is a Lua based module system that easily handles the MODULEPATH Hierarchical problem. Environment Modules provide a convenient way to dynamically change the users’ environment through modulefiles. This includes easily adding or removing directories to the PATH environment variable. Modulefiles for Library packages provide environment variables that specify where the library and header files can be found.
About
Slurm in Docker - Exploring Slurm using CentOS 7 based Docker images
Topics
Resources
License
Uh oh!
There was an error while loading.Please reload this page.
Stars
Watchers
Forks
Releases
No releases published
Packages0
No packages published
Uh oh!
There was an error while loading.Please reload this page.
Contributors4
Uh oh!
There was an error while loading.Please reload this page.