<div dir="ltr">Oh, Maybe the Jacobian has Nans or Infs even though the last time step survived. Maybe it was going crazy. I'll check</div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Nov 1, 2017 at 2:13 PM, Mark Adams <span dir="ltr"><<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Yea, I don't understand the linear solve error:<div><br></div><div>-ts_monitor -ts_type beuler -pc_type lu -pc_factor_mat_solver_package mumps -ksp_type preonly -snes_monitor -snes_rtol 1.e-10 -snes_stol 1.e-10 -snes_converged_reason -snes_atol 1.e-18 -snes_converged_reason<br></div><div><br></div><div>Maybe mumps?</div></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Nov 1, 2017 at 2:02 PM, Jed Brown <span dir="ltr"><<a href="mailto:jed@jedbrown.org" target="_blank">jed@jedbrown.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="m_4001402480670605267HOEnZb"><div class="m_4001402480670605267h5">Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> writes:<br>
<br>
> On Wed, Nov 1, 2017 at 1:46 PM, Jed Brown <<a href="mailto:jed@jedbrown.org" target="_blank">jed@jedbrown.org</a>> wrote:<br>
><br>
>> Mark Adams <<a href="mailto:mfadams@lbl.gov" target="_blank">mfadams@lbl.gov</a>> writes:<br>
>><br>
>> > I have added some code in a TS post step method to look at the number of<br>
>> > nonlinear iterations and cut the time step if it took too many SNES<br>
>> > iterations. That helped but now I want to go one step further and<br>
>> recover<br>
>> > from a failed time step (see appended error message).<br>
>> ><br>
>> > How can/should I recover from a SNES failure inside of TS? I could just<br>
>> put<br>
>> > TSSolve in a loop and if TS failed because of the solver, then reduce the<br>
>> > time step and try to recover the state, and try again.<br>
>> ><br>
>> > Thoughts?<br>
>> ><br>
>> > Thanks,<br>
>> > Mark<br>
>> ><br>
>> > Note, this is a direct solver and the linear solver is diverging? I've<br>
>> > seen the same thing with the maximum its in SNES and failed linesearch.<br>
>> ><br>
>> > 23 TS dt 0.0025 time 0.1025<br>
>> > 0 SNES Function norm 9.997473571601e+04<br>
>> > 1 SNES Function norm 2.295851692533e-01<br>
>> > Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE<br>
>> iterations 1<br>
>> > [0]PETSC ERROR: --------------------- Error Message<br>
>> > ------------------------------<wbr>------------------------------<wbr>--<br>
>> > [0]PETSC ERROR:<br>
>> > [0]PETSC ERROR: TSStep has failed due to DIVERGED_NONLINEAR_SOLVE,<br>
>> increase<br>
>> > -ts_max_snes_failures or make negative to attempt recovery<br>
>><br>
>> Did you increase -ts_max_snes_failures (TSSetMaxSNESFailures)?<br>
>><br>
><br>
> Ah, read the error message :) with a negative value I get this.<br>
><br>
> TS can not, or is not going to try, to reconstruct my state and reduce the<br>
> time step. So I think I need to do this myself.<br>
<br>
</div></div>I don't understand. Why is the linear solve diverging<br>
(-ksp_converged_reason)? What TS are you using?<br>
<div class="m_4001402480670605267HOEnZb"><div class="m_4001402480670605267h5"><br>
> I guess just making loop and testing in the TSConvergedReason, is the way<br>
> to go. We already do something like this for mesh adpativity.<br>
><br>
> Thanks,<br>
><br>
> 23 TS dt 0.0025 time 0.1025<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> 1 SNES Function norm 2.295851692533e-01<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 1<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> 0 SNES Function norm 9.997473571601e+04<br>
> Nonlinear solve did not converge due to DIVERGED_LINEAR_SOLVE iterations 0<br>
> [0]PETSC ERROR: --------------------- Error Message<br>
> ------------------------------<wbr>------------------------------<wbr>--<br>
> [0]PETSC ERROR:<br>
> [0]PETSC ERROR: TSStep has failed due to DIVERGED_STEP_REJECTED<br>
> [0]PETSC ERROR: See <a href="http://www.mcs.anl.gov/petsc/documentation/faq.html" rel="noreferrer" target="_blank">http://www.mcs.anl.gov/petsc/d<wbr>ocumentation/faq.html</a> for<br>
> trouble shooting.<br>
> [0]PETSC ERROR: Petsc Development GIT revision: v3.8-53-ga2d93b2d43 GIT<br>
> Date: 2017-10-09 12:56:49 -0400<br>
> [0]PETSC ERROR: ./ex48 on a arch-macosx-gnu-O named MarksMac-5.local by<br>
> markadams Wed Nov 1 13:51:24 2017<br>
> [0]PETSC ERROR: Configure options --with-cc=clang --with-cc++=clang++<br>
> COPTFLAGS="-O2 -g -mavx2" CXXOPTFLAGS="-O2 -g -mavx2" FOPTFLAGS="-O2 -g<br>
> -mavx2" --download-mpich=1 --download-parmetis=1 --download-metis=1<br>
> --download-hypre=1 --download-ml=1 --download-triangle=1<br>
> --download-ctetgen=1 --download-p4est=1 --download-fftw --download-mumps=1<br>
> --download-scalapack=1 --with-x=0 --download-superlu_dist<br>
> --download-superlu --download-ctetgen --with-debugging=0 --download-hdf5=1<br>
> PETSC_ARCH=arch-macosx-gnu-O --download-chaco --with-viewfromoptions=1<br>
> [0]PETSC ERROR: #1 TSStep() line 4130 in<br>
> /Users/markadams/Codes/petsc/s<wbr>rc/ts/interface/ts.c<br>
> [0]PETSC ERROR: #2 TSSolve() line 4374 in<br>
> /Users/markadams/Codes/petsc/s<wbr>rc/ts/interface/ts.c<br>
> *********** DIVERGED_STEP_REJECTED ierr=91 reason=-2<br>
> L_2 Diff: 1.32241e+07 at time 0.1025<br>
><br>
><br>
><br>
><br>
>><br>
>> > [0]PETSC ERROR: See <a href="http://www.mcs.anl.gov/petsc/documentation/faq.html" rel="noreferrer" target="_blank">http://www.mcs.anl.gov/petsc/d<wbr>ocumentation/faq.html</a><br>
>> for<br>
>> > trouble shooting.<br>
>> > [0]PETSC ERROR: Petsc Development GIT revision: v3.8-53-ga2d93b2d43 GIT<br>
>> > Date: 2017-10-09 12:56:49 -0400<br>
>> > [0]PETSC ERROR: ./ex48 on a arch-macosx-gnu-O named MarksMac-5.local by<br>
>> > markadams Wed Nov 1 13:17:15 2017<br>
>> > [0]PETSC ERROR: Configure options --with-cc=clang --with-cc++=clang++<br>
>> > COPTFLAGS="-O2 -g -mavx2" CXXOPTFLAGS="-O2 -g -mavx2" FOPTFLAGS="-O2 -g<br>
>> > -mavx2" --download-mpich=1 --download-parmetis=1 --download-metis=1<br>
>> > --download-hypre=1 --download-ml=1 --download-triangle=1<br>
>> > --download-ctetgen=1 --download-p4est=1 --download-fftw<br>
>> --download-mumps=1<br>
>> > --download-scalapack=1 --with-x=0 --download-superlu_dist<br>
>> > --download-superlu --download-ctetgen --with-debugging=0<br>
>> --download-hdf5=1<br>
>> > PETSC_ARCH=arch-macosx-gnu-O --download-chaco --with-viewfromoptions=1<br>
>> > [0]PETSC ERROR: #1 TSStep() line 4129 in<br>
>> > /Users/markadams/Codes/petsc/s<wbr>rc/ts/interface/ts.c<br>
>> > [0]PETSC ERROR: #2 TSSolve() line 4374 in<br>
>> > /Users/markadams/Codes/petsc/s<wbr>rc/ts/interface/ts.c<br>
>> > [0]PETSC ERROR: #3 main() line 1161 in /Users/markadams/Codes/mhd/<br>
>> src/ex48.c<br>
>><br>
</div></div></blockquote></div><br></div>
</div></div></blockquote></div><br></div>