본문으로 건너뛰기
This page is not translated yet. 한국어 is switched on so the navigation and URLs can be tested; the text below is the English original. Use the language selector in the header to switch back, or see the documentation home.
how-to · Reviewed 2026-08-27

Troubleshooting

When a job fails, the Rescale platform usually tells you where to look but not always what to do. This page is organized by symptom so you can match what you are seeing to a concrete first step. Start with the general diagnostic pattern, then jump to the section that matches your failure mode.

The general diagnostic pattern​

Whatever the symptom, work in this order. You will narrow most problems to a root cause in a few minutes.

  1. Check the Status tab. Did validation pass with a green check? What does the Gantt chart show? Are there error messages in Job Logs? See The job lifecycle for what each stage means, and Read job logs for which log to open.

  2. Open process_output.log. On the Results page, search for process or log and open it with the screen icon. This file contains stdout from your software, license checkout messages, and most error text. The exit code at the bottom tells you what kind of failure it was:

    Exit codeMeaning
    0No explicit error (may or may not mean correct results)
    1Catchall for general errors
    2Misuse of shell built-ins
    126Command invoked cannot execute
    127Command not found
    137Out of memory or explicit kill
    128+nFatal error signal n

    See Troubleshooting Jobs by Reading process_output.log and Exit Codes on Rescale for a deeper reference.

  3. Check software-specific logs. Most solvers write additional .log, .out, .live, or .dat files with more detail than stdout. For ANSYS, see How to Find Ansys Debug Logs for Troubleshooting and Understanding Ansys License Debug Files.

  4. Check Cluster Metrics. If memory ran out or a node went unhealthy, the root cause is infrastructure, not the input deck.

  5. Re-read the command. Most batch failures are tiny command issues: a wrong path, a missing cd, a Windows line ending, or the wrong software version.

If you need Rescale Support to look at a browser session, capture a HAR file with How to Capture and Export a HAR File for Rescale Support.

Submission and job loading​

Submit button is greyed out​

A required field is missing. Most often it is a required custom field, a missing input file, or an unset license entry. Fill the missing field and the button enables. See Why is the Job Submit Button Greyed Out?.

"We could not load the job that you requested"​

The job page itself fails to load. This is usually a browser state or permissions issue. See "We Could Not Load the Job That You Requested" Troubleshooting Guide.

Sign-in errors​

File, path, and command errors​

Missing input files​

Rescale unpacks your compressed archive into the job's work directory. Confirm the archive contains every file the solver needs and that filenames are correct.

Command not found, or wrong paths​

Most path failures come from how the archive was packaged. Rescale executes the command from the directory where the archive was unpacked, and assumes files are at the top level. If your files are in a subdirectory, prepend cd subdir && to the command, or repackage without the subdirectory. Use relative paths rather than absolute ones.

See Troubleshooting: Common Job Failures on Rescale (Files, Paths, Resources, Licensing).

/bin/bash^M: bad interpreter or similar line-ending errors​

A Windows text editor saved your script with CRLF line endings. Rescale's Linux compute nodes expect LF. Re-save the file as Unix line endings, or clean it in place:

:%s/^M$//

where ^M is Ctrl-V Ctrl-M. See OpenFOAM Error: /bin/bash^M: bad interpreter.

"Command not found" with ampersand in filename​

Special characters like & in filenames confuse the shell. Rename the file. See Troubleshooting Command Not Found Errors with Ampersand Characters in Filenames.

Multi-node worker cannot read files​

For multi-node jobs, workers read from the NFS-mounted ~/work/shared directory. Rescale already defaults most known multi-node solvers to run from ~/work/shared, but if yours does not, prepend:

mv * shared
cd shared

to the command so the input files are reachable on every node.

Resource errors​

Out of memory​

  • Exit code 137 in process_output.log is the classic out-of-memory signature.
  • Avg Free Memory near zero on the Cluster Metrics panel confirms it.

Move to a coretype with more memory per core, reduce the per-core memory footprint, or reduce the input size. For ANSYS CFX, see Resolving Insufficient Memory Allocation Errors in ANSYS CFX.

Disk full​

Pick a coretype with more local disk or reduce the volume of intermediate output your solver writes.

"Detected unhealthy instance(s), shutting down instance group"​

A node in your cluster went unhealthy and Rescale terminated the group to protect the run. This is usually a cloud-provider-level issue. Retry the job, and if it recurs, contact Support. See Troubleshooting: "Detected unhealthy instance(s)".

Job succeeds on X cores but fails on Y cores​

Scaling issues are common in parallel codes: more cores per node means less memory per process, domain decomposition may break at different counts, and some codes hit hardcoded limits. See Troubleshooting: Job succeeds on X cores but fails on Y cores.

Licensing errors​

License failures almost always show up as a checkout error near the top of process_output.log.

For ANSYS license debug files specifically, see Understanding Ansys License Debug Files.

Solver-specific errors​

ANSYS Fluent​

ANSYS CFX​

ANSYS Mechanical​

STAR-CCM+​

Abaqus​

CONVERGE​

Altair OptiStruct​

Cadence​

Stopping and workstation errors​

Job or workstation won't shut down​

Graceful stop sequences run in order: application shutdown, process cleanup, file upload, cluster termination. Any step can hang. If it does, use Force Stop from the Status tab (see stop a job). If force stop also fails, see Troubleshooting: Job or Workstation Won't Shut Down.

Workstation display does not fit the screen​

Resolution negotiation between your client and NICE DCV produced a window that overflows. See Troubleshooting: Workstation Window Doesn't Fit the Screen.

NICE DCV wrong username/password​

See Troubleshooting NICE DCV Wrong Username and Password Error.

Rescale Transfer Manager (RTM)​

Browser file transfers​

Rescale CLI​

Reaching out for help​

When you open a ticket, Rescale Support needs:

  • The job ID (from the ID column on the Jobs list).
  • The text of the relevant error, ideally with a few lines of context from process_output.log.
  • A HAR file if the issue is UI-specific (HAR capture guide).

Sharing this up front shortens the round trip.

On the roadmap​

  • Troubleshooter for Every Job
  • Troubleshooting Agent: Spring Launch

Both will surface recommended next steps inline on the job page for common failure modes.