[ANT]

Hey everyone,

I encountered an odd error today. After running a 22M element case with Fluent v14.5 with the Coupled solver, the case filled the memory (64GB) on my workstation and used about ~15GB of page file. Now, I have the page file on two RAID-0 SSDs, so this didn't slow things down too much, and within ~3-4 hours I had 100 iterations complete. I am running the simulation on a 6-core i7 processor (hyperthreading off). I launched Fluent in 6-process Parallel mode, with Double Precision enabled, from the start (i.e. I didn't run a couple iterations with the serial solver before saving the case and relaunching in the parallel environment). I added a mass-averaged surface monitor for velocity on my outlet surface, which I use to watch convergence. I have run identical simulations before, although on slightly smaller meshes, and did not face this problem.

Then, I closed Fluent as I usually did to get back to Workbench and look at the results. When Fluent is closed, it usually saves the case/data in a compressed file, and then shuts down.

This time, however, I got the following errors during this case-save process (from screenshot, below):


They read (in case screenshot doesn't appear):

Primitive Error at Node 0: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.

Primitive Error at Node 2: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.

Primitive Error at Node 3: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.

Primitive Error at Node 5: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.

Primitive Error at Node 1: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.

Primitive Error at Node 4: Buffered write, in function distribute_random: unable to realloc memory (3181058304 bytes) for mem_buf->mem_buckets.


Does anybody know what these errors mean? Fluent just sat there with the "Saving File" message, and with no processor activity or hard disk activity. Clicking "Cancel" just brought up the "Interrupting Client..." message in the console, but it never managed to "interrupt" it. I had to manually End Process to get Fluent to close. The page file had an extra ~25GB of room available for expansion, but could this error be from the RAM itself being full?

The simulations ran without issue, first a Flow analysis followed by a species transport solution. The problem only came when I tried to save, so I lost the results of the run after having to manually End Process.

I'm trying to run the case again now with the SIMPLE solver rather than the Coupled solver I was using before, in hopes that perhaps lower memory use might help prevent this problem. Any suggestions? I've had no luck in finding any relevant information while Googling or looking through the documentation. I'm also not using any sort of UDFs, etc. Just a plain old Laminar-flow solver on a fluid followed by species transport, using a mesh from ANSYS Meshing.

Thanks in advice for any feedback or assistance!

Edit: After doing the End Process on Fluent and re-launching the Setup cell from Workbench, I tried using the SIMPLE solver rather than the Coupled solver, and this time the RAM didn't fill to capacity and spill over into the page file. However, in addition to taking much longer to run 400 iterations with the SIMPLE solver than it did to run 100 iterations on the Coupled solver (and looking like it needed another ~500 iterations to converge to a similar extent as 100 iterations on the Coupled solver), I still received the identical errors when trying to save the file, and this time only after running the flow simulation, and not the species transport sim that I usually run after.

What error is this, and how do I fix it?

[vasava]

The problem could be that you do not have enough memory to process the data. It is possible that the problem is not with fluent setup but the setup of the computer/cluster that you are using.