AMR Init Fails to Apply to Restart File

[hwhitehead]

Following the routine setup on the AMR page of the wiki, I have a run which requires an initialisation step and a restart.

The format for the setup and restart are as follows

mpiexec -np n ./bin/athena -i athinput.[INPUT] -d [OUTPUT] time/nlim=1

mpiexec -np N ./bin/athena -i [OUTPUT]/nshear.00000.rst -d [OUTPUT] time/nlim=-1

The second call returns an error

### FATAL ERROR in Mesh constructor Too few mesh blocks: nbtotal (180) < nranks (188)

This suggests that the AMR step has not been applied, as the root grid has 180 MeshBlocks. However, when checking the number of MeshBlocks in the *00000.hdf5 file, there appears to be 2210 (as would be expected if AMR was applied properly). I'm not entirely sure how this could be the case, as I would have expected the first .rst and .hdf5 files to be written from the same simulation state.

Is there some other way I should be running the setup/restart routine? Has anyone encountered this error of AMR failing to be saved within the initial .rst file?

This issue ties in somewhat with this recent discussion concerning MPI issues; both these issues could be avoided if MPI ranks were allowed to have zero MeshBlocks. I have no clue how readily this could be implemented, but the whole setup/restart loop could be avoided this way.

[tomidakn]

The second command should be
mpiexec -np N ./bin/athena -r [OUTPUT]/nshear.00000.rst -d [OUTPUT] time/nlim=-1

[hwhitehead]

Good eye, it definitely was -r in previous scripts, must have been a new typo. I'm guessing the .rst files also contains a copy of the athinput state, which it reverts to when using -i?

Either way, this seems like an obvious fix. Thanks!