search

ComputationalRadiationPhysics / picongpu

Performance-Portable Particle-in-Cell Simulations for the Exascale Era :sparkles:
https://picongpu.readthedocs.io
Other
692 stars 218 forks source link

adios-issue "WARN : MPI_AMR method: No OST to use." #3380

Closed paschk31 closed 6 months ago

paschk31 commented 3 years ago

after running a simulation, stderr gives me:

WARN : MPI_AMR method: No OST to use. Set num_ost=NNN in the adios config xml file.

sim ran nevertheless, occured while being on branch (@pordyna): https://github.com/pordyna/picongpu/tree/topic-collisions

I'm not sure where the warning is coming from and how to prevent it/what will happen if i just ignore it

(already disussed with @n01r and @psychocoderHPC )

PrometheusPi commented 3 years ago

@paschk31 The warning is coming from ADIOS/openPMD. What plugins have you activated (openPMD, adios, xrayScattering, etc.)? It does not now, how many "data-sinks" it should use.

n01r commented 3 years ago

I think I remember a conversation with our local admins, saying that we do not actually have OSTs on Hemera but something similar. So I suppose the error comes from our code not being able to use the right "words" to request the OST-like structures.

PrometheusPi commented 3 years ago

Did you get any more errors (as in #2861)?

paschk31 commented 3 years ago

@paschk31 The warning is coming from ADIOS/openPMD. What plugins have you activated (openPMD, adios, xrayScattering, etc.)?

i use these plugins: 

TBG_plugins="!TBG_e_histogram !TBG_H_histogram                                   \
             !TBG_e_PSypy !TBG_H_PSypy                                           \
             !TBG_sumEnergy !TBG_chargeConservation                              \
             !TBG_openPMD"
paschk31 commented 3 years ago

Did you get any more errors (as in #2861)?

no just the one above, 30 times

sbastrakov commented 3 years ago

Not sure, maybe @franzpoeschel has any idea?

franzpoeschel commented 3 years ago

ADIOS1 groups are parameterized by a num_ost parameter (among others) (see p. 63 in the ADIOS user manual) that specifies "the number of Lustre storage targets available in the file system". The openPMD API sets this parameter as 0 by default, hence ADIOS warns that there are no OSTs to use. Look here on how to set the parameter explicitly.

EDIT: Missed a chance to promote ADIOS2 – you may want to checkout the ADIOS2 backend of the openPMD API. This issue will not occur there ;)

BrianMarre commented 6 months ago

@paschk31do you consider this issue still unsolved? If not please close it ;)