[Jay22kar]

Hello foamers,

Recently I considered using AWS for one of my simulations.
I zipped my case directory, copied from my PC to the remote PC, and unzipped it.
As soon as I start the simulation, it starts reading U file & KILLS itself.

The case runs normally in my PC but KILLS itself on the AWS. Please have a look at the screenshot.

The only difference is - my OpenFOAM version is 2006(.com version) & the remote AWS PC contains OpenFOAM V8(.org version)..

PS - Even decomposePar wasn't working before. I copied decomposeParDict file from V8 motorbike tutorial to one of the case & then decomposePar worked!

Has anyone faced such issues? Any solutions?
Thanks

[Tobermory]

Is the data in the U file stored in binary or ASCII? If binary, then will may have problems if the hardware architecture of the 2 platforms (PC, AWS) is different ... write the data files out in ASCII (I seem to remember there is a utility for that?), transfer and then try again (you can switch back to binary writes once you have kicked off the simulations again).

[Jay22kar]

Hello Tobermory,

It's a simple U file containing patches and BCs- which is default in ASCII version, right?

I tried running the case on my friend's PC which had OpenFOAM V8 installed. And it did run! So version is not the culprit I guess.

As far as the binary-ASCII type is concerned, I uploaded another 2D case from my PC, which also worked on AWS! So, if that was the issue, even the 2D case should not have run, right?
Tell me what do you think?

Also, I am not "very" familiar with ubuntu systems, I am still learning, so I would really appreciate if you shed more light on it or any utilities etc.
Thanks

[Tobermory]

Originally Posted by Jay22kar

Hello Tobermory,

It's a simple U file containing patches and BCs- which is default in ASCII version, right?

I tried running the case on my friend's PC which had OpenFOAM V8 installed. And it did run! So version is not the culprit I guess.

As far as the binary-ASCII type is concerned, I uploaded another 2D case from my PC, which also worked on AWS! So, if that was the issue, even the 2D case should not have run, right?
Tell me what do you think?

Well, this is not entirely the case. Take a look at a binary boundary file and you will find that the header is ASCII, and sois the bulk of the boundary data at the end - ie you can read it in a text editor. The binary parts of the file are the lists of internalField data and boundary patch data ... UNLESS they are simple, eg zeroGradient boundary condition or uniform value across the patch or field, in which case the entry will still be in ASCII.

So your 2D example doesn't rule out binary incompatibility as an issue. Have a quick look at the 2D case boundary files - are they all ASCII? Now look at your full case example - are parts of the files in binary (eg the values appear garbled in a text editor)? Probably yes.

The utility I was talking about is foamFormatConvert - that allows you to rewrite a binary file as ASCII, and can help when porting data from one machine type to another. The ASCII file will be big, ofc, so you probably want to gzip it before transferring across.