tree: bbe0eee0cdd17e6a91f9da30deade852a2ac7e99 [path history] [tgz]
  1. Makefile.am
  2. Makefile.in
  3. README.md
  4. slurm.vim
  5. slurm_completion.sh
contribs/slurm_completion_help/README.md

slurm-helper

Bunch of helper files for the Slurm resource manager

Vim syntax file

The Vim syntax file renders the Slurm batch submission scripts easier to read and to spot errors in the submission options.

As submission scripts are indeed shell scripts, and all Slurm options are actually Shell comments, it can be difficult to spot errors in the options.

This syntax file allows vim to understand the Slurm option and highlight them accordingly. Whenever possible, the syntax rules check the validity of the options and put in a special color what is not recognized as a valid option, or valid parameters values.

Installation

Under Linux or MacOS, simply copy the file in the directory

.vim/after/syntax/sh/

or whatever shell other than sh you prefer.

The syntax file is then read and applied on a Shell script after the usual syntax file has been processed.

Known issues

  • Some regex needed to validate options or parameter values are not exactly correct, but should work in most cases.
    • Currently, value completions do not support complex types (e.g. param=key0:val0,key1:val1,).
  • Any new option unknown to the syntax file will be spotted as an error.
  • On a Debian system (Ubuntu) you may see messages like... _get_comp_words_by_ref: command not found after a tab. Based on http://askubuntu.com/questions/33440/tab-completion-doesnt-work-for-commands you need to alter your /etc/bash.bashrc to make this work correctly.

Bash completion

The Bash completion script offers completion for Slurm commands.

Installation

You must have the bash-completion package installed for the script to work properly.

There are a number of ways to configure and install bash completions for Slurm. Please refer to bash-completion for general information.

Enable for all users:

  • Install Slurm contribs:
    ./configure
    sudo make install-contrib
    
  • Install just contribs/slurm_completion_help
    ./configure
    cd contribs/slurm_completion_help
    sudo make install
    
  • Build and install certain Slurm from RPM or DEB packages.
    • DEB: install slurm-smd-client_[0-9]*.deb
    • RPM: install slurm_[0-9]*.rpm

Enable for one user:

  • Copy slurm_completion.sh into $HOME/.local/share/bash-completion/completions, and symbolic link each supported Slurm command to slurm_completion.sh:
    for cmd in sacct sacctmgr salloc sattach sbatch sbcast scancel scontrol scrontab sdiag sinfo slurmrestd sprio squeue sreport srun sshare sstat strigger; do
      ln -sf "$HOME"/.local/share/bash-completion/completions/{slurm_completion.sh,"$cmd"};
    done
    
  • Copy slurm_completion.sh into $HOME, and add a line to source it in $HOME/.bashrc:
    cp contribs/slurm_completion_help/slurm_completion.sh $HOME/slurm_completion.sh
    echo ". $HOME/slurm_completion.sh" >>~/.bashrc
    

Additionally, there are a number of environment variables that can be set to alter/customize completion behavior. They are documented in slurm_completion.sh.

Examples

$ sacct --<tab><tab>
--accounts=       --endtime=        --local           --state=
--allclusters     --env-vars        --long            --timelimit-max=
--allocations     --federation      --name=           --timelimit-min=
--allusers        --fields=         --ncpus=          --truncate
--associations=   --file=           --nnodes=         --uid=
--autocomplete=   --flags=          --noconvert       --units=
--batch-script    --format=         --nodelist=       --usage
--brief           --gid=            --noheader        --use-local-uid
--cluster=        --group=          --parsable        --user=
--clusters=       --help            --parsable2       --verbose
--completion      --help-fields     --partition=      --version
--constraints=    --helpformat      --qos=            --wckeys=
--delimiter=      --jobs=           --reason=         --whole-hetjob=
--duplicates      --json            --starttime=      --yaml
$ squeue --<tab><tab>
--accounts=      --hide           --nodelist=      --states=
--all            --iterate=       --nodes=         --steps=
--array          --jobs=          --noheader       --usage
--array-unique   --json           --partitions=    --user=
--autocomplete=  --licenses=      --priority       --users=
--cluster=       --local          --qos=           --verbose
--clusters=      --long           --reservation=   --version
--federation     --me             --sib            --yaml
--format=        --name=          --sibling
--Format=        --noconvert      --sort=
--help           --node=          --start

$ squeue --users=<tab><tab>
root,    slurm,    user0,    user1,    user2,    user3,

$ squeue --Format=<tab><tab>
Display all 118 possibilities? (y or n)
account,            licenses,           resvport,
accruetime,         maxcpus,            schednodes,
admin_comment,      maxnodes,           sct,
allocnodes,         mcslabel,           selectjobinfo,
allocsid,           mem-per-tres,       siblingsactive,
arrayjobid,         mincpus,            siblingsactiveraw,
arraytaskid,        minmemory,          siblingsviable,
associd,            mintime,            siblingsviableraw,
batchflag,          mintmpdisk,         sockets,
batchhost,          name,               sperboard,
boardspernode,      network,            starttime,
burstbuffer,        nice,               state,
burstbufferstate,   nodelist,           statecompact,
cluster,            nodes,              stderr,
clusterfeature,     ntperboard,         stdin,
command,            ntpercore,          stdout,
comment,            ntpernode,          stepid,
container,          ntpersocket,        stepname,
contiguous,         numcpus,            stepstate,
cores,              numnodes,           submittime,
corespec,           numtasks,           system_comment,
cpufreq,            origin,             threads,
cpus-per-task,      originraw,          timeleft,
--More--
$ scontrol <tab><tab>
abort               pidinfo             shutdown
cancel_reboot       ping                suspend
cluster             reboot              takeover
completing          reconfigure         token
create              release             top
delete              requeue             uhold
errnumstr           requeuehold         update
fsdampeningfactor   resume              version
help                schedloglevel       wait_job
hold                setdebug            write
listpids            setdebugflags
notify              show

$ scontrol update <tab><tab>
frontendname=     nodename=         reservationname=
jobid=            partitionname=    stepid=

$ scontrol update nodename=node<tab><tab>
node00,  node03,  node06,  node09,  node12,  node15,  node18,
node01,  node04,  node07,  node10,  node13,  node16,  node19,
node02,  node05,  node08,  node11,  node14,  node17,

$ scontrol update nodename=node00 <tab><tab>
activefeatures=     extra=              nodename=
availablefeatures=  gres=               reason=
comment=            nodeaddr=           state=
cpubind=            nodehostname=       weight=

$ scontrol update nodename=node00 state=<tab><tab>
cancel_reboot      future             power_down_asap    undrain
down               idle               power_down_force
drain              noresp             power_up
fail               power_down         resume