Interactive Jobs

An interactive job gives you a shell on a compute node where you can type commands live for hands-on, exploratory work. It's the right tool for testing a smaller version of a job before you scale up, debugging code, benchmarking, or working through a tutorial for new software. The steps below assumes that you have successfully ssh into a login node using a terminal.

Starting an interactive session

The easy way: crc-interactive

The crc-interactive command builds the Slurm request for you. The simplest invocation grabs the defaults — one core on one SMP node for one hour:

crc-interactive -s

Add flags to request more (cluster, cores, memory, time, GPUs). Pass the command the --help flag to display the full list of options

commandoutput

crc-interactive --help

usage: crc-interactive [-h] [-v] [-z] [-p PARTITION] [-s] [-g] [-m] [-i] [-d] [-e] [-b MEM] [-t TIME] [-n NUM_NODES] [-c NUM_CORES]
                       [-u NUM_GPUS] [-a ACCOUNT] [-r RESERVATION] [-l LICENSE] [-f FEATURE] [-o]

Launch an interactive Slurm session.

optional arguments:
  -h, --help                                 show this help message and exit
  -v, --version                              show program's version number and exit
  -z, --print-command                        print the equivalent slurm command and exit

Cluster Arguments:
  -p PARTITION, --partition PARTITION        run the session on a specific partition
  -s, --smp                                  launch a session on the smp cluster
  -g, --gpu                                  launch a session on the gpu cluster
  -m, --mpi                                  launch a session on the mpi cluster
  -i, --invest                               launch a session on the invest cluster
  -d, --htc                                  launch a session on the htc cluster
  -e, --teach                                launch a session on the teach cluster

Arguments for Increased Resources:
  -b MEM, --mem MEM                          memory in GB
  -t TIME, --time TIME                       run time in hours or hours:minutes [default: 01:00:00]
  -n NUM_NODES, --num-nodes NUM_NODES        number of nodes [default: 1]
  -c NUM_CORES, --num-cores NUM_CORES        number of cores per node [default: 1]
  -u NUM_GPUS, --num-gpus NUM_GPUS           if using -g, the number of GPUs [default: 0]

Additional Job Settings:
  -a ACCOUNT, --account ACCOUNT              specify a non-default account
  -r RESERVATION, --reservation RESERVATION  specify a reservation name
  -l LICENSE, --license LICENSE              specify a license
  -f FEATURE, --feature FEATURE              specify a feature, e.g. `ti` for GPUs
  -o, --openmp                               run using OpenMP style submission

Tip

Add -z to any crc-interactive command to print the equivalent srun command without running it.

commandoutput

crc-interactive -s -z

srun -M smp --export=ALL --nodes=1 --time=01:00:00 --mem=1g --ntasks-per-node=1 --pty bash

The underlying command: srun

crc-interactive ultimately calls srun, which you can also use directly:

[gnowmik@login1 ~]$ srun -M smp --export=ALL --nodes=1 --time=01:00:00 --mem=1g --ntasks-per-node=1 --pty bash
srun: You have specified NO/WRONG partition, so defaulting to the smp partition.
srun: job 23650051 queued and waiting for resources
srun: job 23650051 has been allocated resources
[gnowmik@smp-n215 ~]$

This commmand asks the scheduler for a session on the smp cluster (-M) with one core (--ntasks-per-node) and 1GB of RAM (--mem) on one node (--nodes) for one hours (--time), running bash in terminal mode (--pty).

When the session starts, your prompt changes from a login node (login1) to the assigned compute node (smp-n215).

Interactive jobs cost Service Units

An interactive session draws Service Units from your group's Resource Allocation for as long as it's held. Be sure to exit out of the compute node when you're done.

If you belong to multiple Resource Allocations, you can select which one to charge against by explicitly specifying --account=<group>. To see your default, run

sacctmgr show associations onlydefaults format=cluster,account%30s,user | grep $USER

Remove the onlydefaults option to show all your Resource Allocations.

GUI applications (X11 forwarding)

To run a graphical application from the terminal, you need enable X11 forwarding from a compute node. First, allocate a node with the salloc command, then ssh into it with -X flag:

[gnowmik@login1 ~]$ salloc -M htc --nodes=1 --time=02:00:00
salloc: You have specified NO/WRONG partition, so defaulting to the htc partition.
salloc: Granted job allocation 10721347
salloc: Waiting for resource configuration
salloc: Nodes htc-n72 are ready for job
[gnowmik@login1 ~]$

[gnowmik@login1 ~]$ ssh -X htc-n72
The authenticity of host 'htc-n72 (10.201.8.72)' can't be established.
ED25519 key fingerprint is SHA256:LSRI8iKoJTxzJprwTWQkr55XuZ/UtTYlk5AtS04Icz0.
This key is not known by any other names
Are you sure you want to continue connecting (yes/no/[fingerprint])? yes
Warning: Permanently added 'htc-n72' (ED25519) to the list of known hosts.
[gnowmik@htc-n72 ~]$
[gnowmik@htc-n72 ~]$ module load xeyes
[gnowmik@htc-n72 ~]$ xeyes

Once connected, load a GUI application with the module system and launch it from the command line.

Note

The ssh -X <compute_node> hop from a login node to your compute node relies on passwordless SSH within the cluster. If you're prompted for a password or denied, set it up following Passwordless SSH → between cluster nodes.

Remember to scancel any interactive job that was started with salloc. Otherwise, the allocated resources remain held until completion of the requested walltime limit. As shown below, the job remains in the queue in the Running state even after I logged out of the compute node.

[gnowmik@htc-n72 ~]$ module load xeyes
[gnowmik@htc-n72 ~]$ xeyes

^C
[gnowmik@htc-n72 ~]$ exit
logout
Connection to htc-n72 closed.
[gnowmik@login1 ~]$ squeue -M htc -u $USER
CLUSTER: htc
             JOBID PARTITION     NAME     USER ST       TIME  NODES NODELIST(REASON)
          10721347       htc interact  gnowmik  R       5:15      1 htc-n72
[gnowmik@login1 ~]$

The scontrol command shows more details about the allocated interactive job that was spawned with salloc -M htc --nodes=1 --time=02:00:00. The --nodes=1 flag allocated one CPU core on a single node along with the default 8GB of RAM per core for that particular node.

[gnowmik@login1 ~]$ scontrol -M htc show job 10721347
JobId=10721347 JobName=interactive
   UserId=gnowmik(152289) GroupId=kwong(16260) MCS_label=N/A
   Priority=14105 Nice=0 Account=kwong QOS=htc-htc-s
   JobState=RUNNING Reason=None Dependency=(null)
   Requeue=1 Restarts=0 BatchFlag=0 Reboot=0 ExitCode=0:0
   RunTime=00:13:52 TimeLimit=02:00:00 TimeMin=N/A
   SubmitTime=2026-08-02T20:10:30 EligibleTime=2026-08-02T20:10:30
   AccrueTime=Unknown
   StartTime=2026-08-02T20:10:30 EndTime=2026-08-02T22:10:30 Deadline=N/A
   PreemptEligibleTime=2026-08-02T20:10:30 PreemptTime=None
   SuspendTime=None SecsPreSuspend=0 LastSchedEval=2026-08-02T20:10:30 Scheduler=Main
   Partition=htc AllocNode:Sid=login1:400755
   ReqNodeList=(null) ExcNodeList=(null)
   NodeList=htc-n72
   BatchHost=htc-n72
   NumNodes=1 NumCPUs=1 NumTasks=1 CPUs/Task=1 ReqB:S:C:T=0:0:*:*
   ReqTRES=cpu=1,mem=8000M,node=1,billing=1
   AllocTRES=cpu=1,mem=8000M,node=1,billing=1
   Socks/Node=* NtasksPerN:B:S:C=0:0:*:* CoreSpec=*
   MinCPUsNode=1 MinMemoryCPU=8000M MinTmpDiskNode=0
   Features=(null) DelayBoot=00:00:00
   OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)
   Command=/bin/bash
   WorkDir=/ihome/kwong/gnowmik
   Power=

[gnowmik@login1 ~]$

If you need more, 67 cores for example, you can pass to salloc the additional flag --ntasks-per-node=67, and Slurm will schedule the job

commandoutput

[gnowmik@login1 ~]$ salloc -M htc --nodes=1 --time=02:00:00 --ntasks-per-node=67
salloc: You have specified NO/WRONG partition, so defaulting to the htc partition.
salloc: Granted job allocation 10722043
salloc: Nodes htc-n87 are ready for job
[gnowmik@login1 ~]$ scontrol -M htc show job 10722043

JobId=10722043 JobName=interactive
   UserId=gnowmik(152289) GroupId=kwong(16260) MCS_label=N/A
   Priority=14084 Nice=0 Account=kwong QOS=htc-htc-s
   JobState=RUNNING Reason=None Dependency=(null)
   Requeue=1 Restarts=0 BatchFlag=0 Reboot=0 ExitCode=0:0
   RunTime=00:00:18 TimeLimit=02:00:00 TimeMin=N/A
   SubmitTime=2026-08-03T00:50:00 EligibleTime=2026-08-03T00:50:00
   AccrueTime=Unknown
   StartTime=2026-08-03T00:50:00 EndTime=2026-08-03T02:50:00 Deadline=N/A
   PreemptEligibleTime=2026-08-03T00:50:00 PreemptTime=None
   SuspendTime=None SecsPreSuspend=0 LastSchedEval=2026-08-03T00:50:00 Scheduler=Main
   Partition=htc AllocNode:Sid=login1:370003
   ReqNodeList=(null) ExcNodeList=(null)
   NodeList=htc-n87
   BatchHost=htc-n87
   NumNodes=1 NumCPUs=67 NumTasks=67 CPUs/Task=1 ReqB:S:C:T=0:0:*:*
   ReqTRES=cpu=67,mem=536000M,node=1,billing=67
   AllocTRES=cpu=67,mem=536000M,node=1,billing=67
   Socks/Node=* NtasksPerN:B:S:C=67:0:*:* CoreSpec=*
   MinCPUsNode=67 MinMemoryCPU=8000M MinTmpDiskNode=0
   Features=(null) DelayBoot=00:00:00
   OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)
   Command=/bin/bash
   WorkDir=/ihome/kwong/gnowmik
   Power=

You see in the output to scontrol that the job was allocated NumNodes=1 NumCPUs=67 NumTasks=67 and with mem=536000M, which is the aggregate RAM for 67 cores at 8GB/core.

GUI applications (Open OnDemand and Viz)

Open OnDemand runs common GUI tools — RStudio, Jupyter Notebook, JupyterLab, and MATLAB — in your browser. It's the easiest option if you're unfamiliar with Slurm or the command line.

Alternatively, for a full graphical Linux desktop, the VIZ portal gives you GUI access using a web browser.

Quality of Service (QoS)

Every job — interactive or batch — is assigned a Quality of Service that caps its walltime and influences its scheduling priority. The default is normal; request another with --qos=<name>. A job asking for more walltime than its QoS allows will be rejected.

See Job Limits & QoS for current limits

The QoS levels, their maximum walltimes and priority factors, and the related per-group CPU/GPU and memory limits are maintained in the Job Limits & QoS page.

Where to go next

• Run work unattended

  ---

  Turn a tested command into a batch script that runs without you.

  Batch Jobs

• Understand the cost

  ---

  How Service Units are calculated and charged against your allocation.

  Service Units

• Work in the browser

  ---

  Notebooks, RStudio, MATLAB, and desktops without command-line setup.

  Open OnDemand

---