LS-DYNA runs into Segmentation Violation (SIGSEGV) error. What can I do?
Content:
.
Error message (Segmentation Violation )
>>>>> Process 18 <<<<<
>>>>> Signal 11 : Segmentation Violation <<<<<
.
Possible procedure for the debugging of the LS-DYNA Error:
1. Is it possible to reproduce the LS-DYNA Error?
- Run the identical input with identical configurations
- For MPP Run: extract NODOUT from the BINOUT* via l2a
Compare the two runs: diff between both NODOUT and both GLSTAT files
--> No comparability: Hardware problem? (cpu, memory, disk, network, ...)
.
2. Identify the LS-DYNA Error
- activate checknan option (*CONTROL_SOLUTION, ISNAN=1)
- Check the energy progression (GLSTAT, MATSUM, if necessary SLEOUT)
- Can irregularities be detected immediately before the termination?
- Which structural parts are affected?
- Do elements fail immediately before the termination?
- check-failed –l mes*
- check-failed –pid –l mes* (Which PIDs are affected?)
- Which are the last messages in STDOUT / STDERR / mes00xx?
- Are there any problems with viscoplasticity (VP=1, “…did not converge”)?
- check-hsp d3hsp –xcon mes*
- Have unreasonable contact settings been applied? (e.g. large penalties- or thickness scaling etc.)
- check-hsp d3hsp –cfac
- Can initial penetrations be detected? How many? How large?
- e.g. check-c mes* -buc –typ –typ
- Are there any problems with spot welds or tied contacts? (e.g. untied nodes = cold spot welds)
- e.g. check-c mes* -select tied
- Within which phase the error occurs?
- Keyword processing, initialization, decomposition 1-3, solution
.
3. Debug-Run
- Switch-off „Plotcprs“ (velocity, stress, strain … available)
- Termination Cycle = <Error Cycle> - 1
- investigate last plot state w.r.t. max. velocities, max. plastic strains, etc..
.
.
Checknan option (Segmentation Violation)
How can I activate the checknan option?
- Command-Line: checknan=1
- Keyword: *CONTROL_SOLUTION, ISNAN=1
.
Which output is produced by checknan option?
- Output to screen (e.g. jobout (STDOUT)):
*** NaN detected. Please check message file from processor for detail.
*** Error NaN detected on processor # 30 <<<
- Output to messag-files (e.g. grep ‘out-of-range‘ mes0030):
*** termination due to out-of-range forces
number of nodes has out-of-range forces 364
Node list:
37502068 37502162 37502192 37502206 37502357 ....
.
Visualization of output by checknan option (nodes)?
- DYNA Tool: check-failed –xnan –a4nan
- Animator: Read in the Session-File ´fail-ide.ses´
.
What does NaN mean?
- NaN („Not a Number“): http://en.wikipedia.org/wiki/NaN
- Node forces/ -moments/ -velocities with NaN occur due to different reasons – e.g. division by zero
.
check-hsp, check-failed, check-c, plotcprs are DYNAmore tools (free for customers: http://www.dynamore.de/tools)
.
.
Error message (MPI)
MPI Application rank 0 exited before MPI_Finalize() with status 13
forrtl: error (78): process killed (SIGTERM)
- „…MPI_Finalize() with status 13“ is an MPI message due to a possible LS-DYNA Error
- Is not important for debugging of the LS-DYNA error.
gp 04/14