assert (!closed) failed - error waiting for event

[Mark]

Hi, all

I've been getting this error mid run pretty recently, on different models as well.

FDS is up to date and using fds_local,

any help will be appreciated thank you.

[Randy McDermott]

When I Google the first error "proxy_cb.c (68): assert (!closed) failed" it appears related to MPI.  Please show exactly the command you used to launch the job and what platform you (Windows, Linux) you are using (looks like Windows, but want to be sure).  Thanks

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/ac2d30ea-82e0-4f8a-9db1-39b1d7bb8628%40googlegroups.com.

[Mark]

fds_local -p 12 -o 1 (filename).fds

Window 10

[Glenn Forney]

Do you have pyrosim installed and/or are you running any other fds cases?

On Wed, Apr 8, 2020, 7:50 AM Mark <markskth...@gmail.com> wrote:

fds_local -p 12 -o 1 (filename).fds

Window 10

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/0fad12ff-a321-40f3-a072-f13f036f1e71%40googlegroups.com.

[Glenn Forney]

Also , how long did it take for your case to get to the point where it failed? 

On Wed, Apr 8, 2020, 7:50 AM Mark <markskth...@gmail.com> wrote:

fds_local -p 12 -o 1 (filename).fds

Window 10

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/0fad12ff-a321-40f3-a072-f13f036f1e71%40googlegroups.com.

[Mark]

Pyrosim is installed on the computer but the model is run through CMDfds, this particular model ran until failure at 552s whilst another model with the same error failed at around 200s

[Mark]

And, only one model is being ran at a time

[Glenn Forney]

Open  up a cmdfds shell. Type where fds then type where mpiexec

On Wed, Apr 8, 2020, 8:25 AM Mark <markskth...@gmail.com> wrote:

Pyrosim is installed on the computer but the model is run through CMDfds, this particular model ran until failure at 552s whilst another model with the same error failed at around 200s

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/ef1083a9-ff91-42a2-88d9-a482252a6e64%40googlegroups.com.

[Kevin]

Attach your input file and we'll run it on one of our computers.

[Mark]

I've sent it as a private message as the project is confidential. 

Thank you :) 

[o...@aquacoustics.biz]

Hi there Mark.

This sounds like an Hydra MPI issue associated with allocated processes (hence the proxy).

If your model is commercial then I can understand your reluctance to share it.  But I'll happily run your model with complete confidentiality on FireNZE's cluster to help resolve your issue if this will help.

As a suggested diagnostic you might want to re-run the model with the same allocated resources and see if you get the same failure at the same time.  If the time or failure mode change then this may suggest a hardware issue.

With kindest regards,

t. http://www.fire.aquacoustics.biz/

e then the Aprocesses allocated under 

[Kevin]

I started the job on a Windows PC with 64 GB of memory and I started it also on a single node of a linux cluster with 12 cores and 64 GB RAM. On the linux computer, the job uses 36 GB RAM and is running slow. On my Windows PC, the job was so big that it just made the machine freeze and I lost my remote connection to it. I finally just had to reboot the thing remotely. 

I cannot say exactly what the issue is, but if these jobs fail at different points in the simulation, then it probably involves memory usage or other hardware issue. This is not the kind of job that I would run on a single Windows computer. The OS is just not built for it. I would break this case up into smaller meshes and run on a dedicated linux cluster.

[Mark]

Thank you Kevin, I appreciate your time looking at this :)