Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Hello! In this section we'll talk about the post processing module and the data visualization.
badawika
Posts: 6
Joined: Thu Sep 29, 2022 8:29 pm

Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by badawika » Thu Sep 29, 2022 9:39 pm

Dear @ Support Team,

I am facing a problem postprocessing the MPCO recorder container after solving a model using OpenseesSP. The output MPCO files are separated into many files according to the running solution processes. When I post-process the model outputs, after finishing the simualtion phase, I find only a portion of the model is plotted, and the other parts are not plotted. Why did this issue appear, and how to solve it or combine the MPCO hdf5 files into only one file to overcome this issue?

I really appreciate any help you can provide.
Attachments
Screenshot 2022-09-30 053415.png
An output contour of two processes model solution
Screenshot 2022-09-30 053415.png (130.39 KiB) Viewed 2760 times

STKO Team
Posts: 2321
Joined: Tue Oct 29, 2019 8:45 am

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by STKO Team » Fri Sep 30, 2022 7:28 am

You should have multiple mpco files, with the same name, and a postfix like this:
SP.png
SP.png (10.02 KiB) Viewed 2757 times
Can you confirm?

badawika
Posts: 6
Joined: Thu Sep 29, 2022 8:29 pm

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by badawika » Fri Sep 30, 2022 12:47 pm

I have only one file with the extension "mpco.data" and many files with the extension ".mpco". I do not know why only one file with "mpco.data" is available; how to solve this issue?
Attachments
Screenshot 2022-09-30 202824.png
Screenshot 2022-09-30 202824.png (12.57 KiB) Viewed 2755 times
Screenshot 2022-09-30 202742.png
Screenshot 2022-09-30 202742.png (55.49 KiB) Viewed 2755 times

STKO Team
Posts: 2321
Joined: Tue Oct 29, 2019 8:45 am

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by STKO Team » Mon Oct 03, 2022 3:03 pm

I have only one file with the extension "mpco.data" and many files with the extension ".mpco"
If you are using OpenSeeSP it's normal that you have only 1 CDATA file.
The error is elsewhere.
When you load one of the MPCO partitioned databases (part-1, part-2, etc..) STKO will try to load all of them. If it finds and error (incompatibility between the partitioned databases) it will write a (RED) error message in the Terminal.
Can you show it? (do a screenshot or copy-paste it here)

Note: sometimes this error happens when you first run OpenSeesSP with N processors. Then you re-run it with M < N processors.
The last (N-M) mpco files are old, and will be incompatible with the new M ones... make sure this is not the case

badawika
Posts: 6
Joined: Thu Sep 29, 2022 8:29 pm

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by badawika » Wed Oct 05, 2022 5:11 pm

This is the output in the terminal window. I have found that this message appeared while opening the output files " Recorder-Sand-Silt-thick-0.3-Dr-34.part-1.mpco has inconsistent number of Model Stages (it will be skipped)".

##########################
End Command: SetContext_PostProcessor
Command: OpenDatabase
Database "F:/NVMe_Data/________Modifed model with CSR total and Pcons Mod/_Void_redist_Models/___Models_silt_Barrier/Infinite_Slopes_D_Safe_PGA_1m_sec2/Sand-DR-76-Silt-0.3m-thick/Recorder-Sand-Silt-thick-0.3-Dr-34.part-0.mpco" correctly open
Database "F:/NVMe_Data/________Modifed model with CSR total and Pcons Mod/_Void_redist_Models/___Models_silt_Barrier/Infinite_Slopes_D_Safe_PGA_1m_sec2/Sand-DR-76-Silt-0.3m-thick/Recorder-Sand-Silt-thick-0.3-Dr-34.part-1.mpco" correctly open
Partition "F:/NVMe_Data/________Modifed model with CSR total and Pcons Mod/_Void_redist_Models/___Models_silt_Barrier/Infinite_Slopes_D_Safe_PGA_1m_sec2/Sand-DR-76-Silt-0.3m-thick/Recorder-Sand-Silt-thick-0.3-Dr-34.part-1.mpco" has inconsistent number of Model Stages (it will be skipped).
Current = 3, next = 2.
Database "F:/NVMe_Data/________Modifed model with CSR total and Pcons Mod/_Void_redist_Models/___Models_silt_Barrier/Infinite_Slopes_D_Safe_PGA_1m_sec2/Sand-DR-76-Silt-0.3m-thick/Recorder-Sand-Silt-thick-0.3-Dr-34.part-1.mpco" correctly closed
End Command: OpenDatabase
Command: SurfaceColorMap
Command: NewPlotGroup
End Command: SurfaceColorMap
#############################

How to solve this issue?

STKO Team
Posts: 2321
Joined: Tue Oct 29, 2019 8:45 am

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by STKO Team » Thu Oct 06, 2022 3:52 pm

This can happen for 2 reasons:

1) OpenSees crashed during the analysis so that one processor wrote more output than another, so the files are inconsistent

2) Some of those MPCO files belong to previous analyses.
May I ask you how many processors are you using?
Do you have number of MPCO files = number of processors?

badawika
Posts: 6
Joined: Thu Sep 29, 2022 8:29 pm

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by badawika » Thu Oct 06, 2022 10:08 pm

All Output MPCO files have similar file sizes and the number of MPCO files equals the number of processes. Also, these simulations have been done on AMD Rzyen 3900X (12 Cores/ 24Threads).

I need to mention that I am running many simulations at the same time and assigning a specific number of processes for each simulation, is that condition affects the MPCO output file structure or result in the corruption of the output files?

Also, sometimes I solve a model considering 4 processes but I stop the analysis and then rerun it again with 8 assigned processes to get the results faster, is that affecting the output structures or not? I delete all output files before rerunning the simulations.

STKO Team
Posts: 2321
Joined: Tue Oct 29, 2019 8:45 am

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by STKO Team » Fri Oct 07, 2022 8:34 am

I need to mention that I am running many simulations at the same time and assigning a specific number of processes for each simulation, is that condition affects the MPCO output file structure or result in the corruption of the output files?
If those simulation are writing to the same output file, then yes, it will be a problem, and the output will be corrupted.
Also, sometimes I solve a model considering 4 processes but I stop the analysis and then rerun it again with 8 assigned processes to get the results faster, is that affecting the output structures or not? I delete all output files before rerunning the simulations.
It's good that you delete the previous MPCO files. However when your new analysis has more processor than the previous one, it's not a problem, because the prevous MPCO will be delete automatically.
The only issue is when the new analysis has less processor. So the remaining files will not be deleted automatically.

Please can you share your SCD files so that we can verify the issue?

badawika
Posts: 6
Joined: Thu Sep 29, 2022 8:29 pm

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by badawika » Thu Oct 13, 2022 8:00 am

Hello, and sorry for my late reply.

Please check the attached file as you requested.

I hope to solve this issue without rerunning the simulations again.

Thanks in advance :).
Attachments
SAND-SAND-test.zip
(217.61 KiB) Downloaded 104 times

STKO Team
Posts: 2321
Joined: Tue Oct 29, 2019 8:45 am

Re: Missed model parts when viewing output contour plot while using OpenSeesSP (only part of the model is viewable)

Post by STKO Team » Thu Oct 13, 2022 1:43 pm

I also have the same issue.
In my case however, when the 3rd stage starts (dynamic) I receive an MPI Error. Probably something that you used in your model does not support (or has some bug) in the sendSelf/recvSelf methods.

Do you have the same error?

Why don't your try using OpenSeesMP? It does not have these issues and with STKO it's very easy to set it up

Post Reply