How to stop YADE from exiting automatically on PBS?

This is the output from YADE, alongside some data files I told it to create. The script prints every 1000th iteration to show it's working during debugging.

------------------------------------------------------
Job is running on node node068
------------------------------------------------------
PBS: qsub is running on login01.cm.cluster
PBS: originating queue is submit
PBS: executing queue is HPCA-02414-PGR
PBS: working directory is /panfs/panasas01.panfs.cluster/evxmk9
PBS: execution mode is PBS_BATCH
PBS: job identifier is 2083995.master.cm.cluster
PBS: job name is gds.sh
PBS: node file is /cm/local/apps/pbspro/var/spool/aux/2083995.master.cm.cluster
PBS: current home directory is /panfs/panasas01.panfs.cluster/evxmk9
PBS: PATH = /cm/shared/apps/gts-0.7.6/gcc/bin:/cm/shared/apps/python/2.7.12/bin:/cm/shared/apps/yade-2$
------------------------------------------------------
TCP python prompt on localhost:9000, auth cookie `escydu'
Welcome to Yade 2016.06a
XMLRPC info provider on http://localhost:21000
Running script gds.py
/cm/shared/apps/python/2.7.12/lib/python2.7/site-packages/IPython/config.py:13: ShimWarning: The `IPyt$
  "You should import from traitlets.config instead.", ShimWarning)
/cm/shared/apps/python/2.7.12/lib/python2.7/site-packages/IPython/core/interactiveshell.py:440: UserWa$
  warn('As of IPython 5.0 `PromptManager` config will have no effect'
1000
2000
3000
[[ ^L clears screen, ^U kills line. F8 plot. ]]

In [1]: Do you really want to exit ([y]/n)?

Hi,

Probably a copy of your YADE python script is as much (at least..) required here as the copy of your PBS script..
Your issue may depend on the kind of commands you're using to let YADE run. For instance O.run(someNumberOfIterations,wait=1) would probably be much better in your case than O.run()

Here is the YADE script.

from yade import pack,bodiesHandling,utils,export
from math import tan,sin,cos,sqrt

internalfrictionangle=28*(pi/180)
externalfrictionangle=20*(pi/180)
width=0.1
rampangle=40*(pi/180)
shutterheight=0.2
radius=0.01
cwidth=width
insertradius=0.2464
shutterpoint=1.448+insertradius*tan(rampangle/2)
ramplength=shutterpoint+shutterheight/tan(rampangle)
runofflength=1.7
generationheight=1.25*cos(rampangle)*shutterheight
damp=0.01

finish=6000000
interval=1000

O.materials.append(FrictMat(density=8050,young=210e6,poisson=0.25,frictionAngle=externalfrictionangle,label='steel'))

#shutter plate(physical)
O.bodies.append(geom.facetBox((cwidth/2,cos(rampangle)*shutterpoint-sin(rampangle)*shutterheight/2,sin(rampangle)*shutterpoint+cos(rampangle)*shutterheight/2),(cwidth/2,shutterheight/2,0), orientation=Quaternion((1,0,0),rampangle-90*(pi/180)),wallMask=63,color=(1,1,1),wire=False))

#shutter wall(for generation)
O.bodies.append(wall((0,cos(rampangle)*shutterpoint-sin(rampangle)*shutterheight,0),axis=1,sense=0,color=(1,1,1)))

O.bodies.append(wall((0,cos(rampangle)*shutterpoint+cos(rampangle)/tan(rampangle)*shutterheight,0),axis=1,sense=0,color=(1,1,1)))

#ramp surface
O.bodies.append(geom.facetBox((cwidth/2,cos(rampangle)*ramplength/2,sin(rampangle)*ramplength/2),(cwidth/2,ramplength/2,0), orientation=Quaternion((1,0,0),rampangle),wallMask=63,color=(1,1,1),wire=False))

#runoff zone
O.bodies.append(geom.facetBox((cwidth/2,-(runofflength+insertradius*tan(rampangle/2))/2,0),(cwidth/2,(runofflength+insertradius*tan(rampangle/2))/2,0),orientation=Quaternion((1,0,0),0),wallMask=63,color=(1,1,1),wire=False))

#curved transition (circular for now)
O.bodies.append(geom.facetCylinder((cwidth/2,-insertradius*tan(rampangle/2),insertradius),insertradius,cwidth,orientation=Quaternion((0,1,0),pi/2),segmentsNumber=16, wallMask=7,angleRange=(0,rampangle),closeGap=False,radiusTopInner=0.999*insertradius,radiusBottomInner=0.999*insertradius,color=(1,1,1),wire=False))

#invisible sidewalls
O.bodies.append(wall((0,0,0),axis=0,sense=0,color=(1,1,1)))
O.bodies.append(wall((cwidth,0,0),axis=0,sense=0,color=(1,1,1)))

O.materials.append(FrictMat(density=1730,young=65e6,poisson=0.2,frictionAngle=internalfrictionangle,label='sand'))

#particle generation
# sphere packing is not equivalent to particles in simulation, it contains only the pure geometry
sp=pack.SpherePack()
# generate randomly spheres with uniform radius distribution
sp.makeCloud((0,cos(rampangle)*shutterpoint-sin(rampangle)*shutterheight,sin(rampangle)*shutterpoint+cos(rampangle)*shutterheight),(cwidth,cos(rampangle)*shutterpoint+cos(rampangle)/tan(rampangle)*shutterheight,generationheight+sin(rampangle)*shutterpoint+cos(rampangle)*shutterheight),rMean=radius,rRelFuzz=0)
# add the sphere pack to the simulation
sp.toSimulation()

O.engines=[
   ForceResetter(),
   InsertionSortCollider([Bo1_Sphere_Aabb(),Bo1_Facet_Aabb(),Bo1_Wall_Aabb()]),
   InteractionLoop(
      [Ig2_Sphere_Sphere_ScGeom(),Ig2_Facet_Sphere_ScGeom(),Ig2_Wall_Sphere_ScGeom()],
      [Ip2_FrictMat_FrictMat_FrictPhys()],
      [Law2_ScGeom_FrictPhys_CundallStrack()]
   ),
   NewtonIntegrator(gravity=(0,0,-9.81),damping=damp),
   PyRunner(command='addData()',iterPeriod=interval),
   PyRunner(command='finish()',iterPeriod=finish),
   PyRunner(command='vtk()',iterPeriod=interval),
]
#O.dt=.5*PWaveTimeStep()
O.dt=5e-7

vtkExporter = export.VTKExporter('/panfs/panasas01.panfs.cluster/evxmk9/YADE/test-data/data')
vtkExporter.exportFacets(what=[('pos','b.state.pos')])

def vtk():
   vtkExporter.exportSpheres(what=[('dist','b.state.pos.norm()'),('particleVelocity','b.state.vel')])

O.run()

def finish():
   O.pause()

#def checkUnbalanced():

#O.trackEnergy=True

def addData():
   print O.iter
   for b in O.bodies:
      if isinstance(b.shape,Sphere): b.shape.color=scalarOnColorScale(b.state.vel.norm(),0,2.5)
   if O.iter==900000:
      for b in O.bodies:
         if isinstance(b.shape,Sphere):
            if b.state.pos[2]>(sin(rampangle)*shutterpoint+cos(rampangle)*shutterheight):
               O.bodies.erase(b.id)
   if O.iter==1000000:
      for b in O.bodies:
         if b.id<4:
            O.bodies.erase(b.id)

O.saveTmp()

Again this code did not exit automatically and functioned normally when I used it on my local PC or on the local server of the HPC (so without using PBS).

I have no or very little idea what HPC or PBS are :-) but you do have a O.run() which directly returns to terminal (with the simulation running in the background), provided the terminal is "active" (like in a "normal" interactive YADE session).

Hence I'm not surprised of the behaviour you're facing, I guess the same would happen with yade-batch

I think adding O.wait() after O.run() may solve your problems without further changes in your script. See
https://yade-dem.org/doc/yade.wrapper.html#yade.wrapper.Omega.wait
and
https://yade-dem.org/doc/user.html#stop-conditions

Jerome

Note that for the same reason your script is weird - at least unpredictable - independently of PBS:

O.run()
def addData():
   [...]
O.saveTmp()

If you are lucky python will read the definition of addData() before the function is effectively used at first iteration, if you are not lucky you may get a "unknown function" error. As for the saveTmp(), nobody can tell if it will save the initial state or the state after 1-2 iterations. For sure, it is not saving the final state.

B

Thanks Bruno Chareyre, that solved my question.

I can confirm that this issue was with O.run(); with a few other minor tweaks the script is now working fine. :D