[Gang Shen]

Hello, everyone! I encountered frequent Fluent exit and cannot resolve this issue for a long time!

I am running parallel CPU-intensively projects in batch mode on higher performance computer clusters using Fluent 18.0. My problem is: Fluent interrupted and exit frequently. E. g., I set Fluent to run 10000 time steps, however, Fluent exit at some time step, for example, at 1000th time step. And I got an error message(which started with ‘Unable to parse: [free’, see below) at the ending of the output file, solution.trn, which contains the information normally printed in the Fluent console window when run at interactive mode.


..............................
Updating solution at time level N... done.
iter continuity x-velocity y-velocity energy k epsilon liquid-fra time/iter
!1904910 solution is converged
1904910 1.1980e-06 1.1668e-11 1.0234e-11 1.4947e-10 1.3142e-16 2.1679e-16 2.8615e-01 0:00:01 30
1904920 3.9598e-04 2.7670e-09 2.0772e-09 3.3919e-10 1.2978e-16 2.2320e-16 2.8615e-01 0:00:00 20
1904930 1.9351e-05 1.9525e-10 1.6919e-10 3.4960e-10 1.3007e-16 2.1505e-16 2.8615e-01 0:00:00 10

Unable to parse: [free

Error: undefined read macro
Error Object: ()

> /file/write-case-data "/scratch/gshen/pro0216/grid500x20x5-0216-%-5.3f"cx-expand-filename: inquire-flow-time error.
cx-expand-filename: inquire-flow-time error.

Error: CAR: invalid argument [1]: wrong type [not a pair]
Error Object: #f


>
Error: wta(1st) to string_ci_eq
Error Object: ()

Error: wta(1st) to string_ci_eq
Error Object: 16384
/parallel/timer/usage

>
Error: wta(1st) to string_ci_eq
Error Object: ()
/parallel/latency

>
Error: wta(1st) to string_ci_eq
Error Object: ()
/parallel/bandwidth

>
Error: lmakevector: wta(1)
Error Object: ()
exit


When I re-launch Fluent from the interrupted point and calculate more time-steps, it would exit again after some time steps running. I had to re-launch again and again to finish a simulation project. This situation is very annoying!
The calculated results are reasonable and the same project could run smoothly on my desktop computer.

The command I submitted to apply for a PBS job on the cluster is,

> qsub -I -l nodes=1 ppn=14,mem=32G,walltime=48:0:0

The command I run Fluent is,

> fluent 2ddp -g -t10 -cflush -i /scratch/myusername/pro0219/pro0219.jou >& /scratch/ myusername /pro0219/solution.trn &

My Fluent journal file, pro0219.jou, contains the followings,

/file/set-batch-options no yes yes no
(set! checkpoint/exit-filename "/scratch/myusername/pro0219/exit-fluent-pro0219.txt")
/file/read-case "/scratch/myusername/pro0216/-182.cas"
/file/read-data "/scratch/myusername/pro0216/-182-10696.790.dat"
/file/auto-save/data-frequency 1000
/file/auto-save/case-frequency/if-case-is-modified
/file/auto-save/root-name "/scratch/myusername/pro0219/"
/file/auto-save/append-file-name-with flow-time 3
/solve/set/time-step 0.1
/solve/dual-time-iterate 10000 30
/file/write-case-data "/scratch/myusername/pro0219/0216-%-5.3f"
/parallel/timer/usage
/parallel/latency
/parallel/bandwidth
exit


Any one has encountered similar situations or could provide some suggestions? Thank you in advance!

[Gang Shen]

Could any suggestion be given?

[Gang Shen]

No one knows this issue?

[Ko otto]

Hi

If you are still in problem, try this;

(1) write bc file from TUI using "/file/write-bc"
(2) find the phrase "cx-virtual-id-list" in the file. and you might find "#[free" (# cells)]".
(3) if you find "#[free" (# cells)]", delete it and re-read bc file using "file/read-bc"

I don't guarantee this works well, but worth to try.
notice that take a backup before try this.

[Gang Shen]

Hi Ko otto,

Thank for your response!

As for the problem I posted, I have not totally solved it but decreased the interruption frequency to the very low level I can accept. My solution is that when you transfer your case and data files from one operating system(e.g. windows) to a different OS(e.g. Linux) to run the computing, you should load the case files in the new OS, modify the auto-saved directories, save and reopen them. Then you go ahead with your computing. I do not know the reasons but by this method the interruption decreases greatly.

I will also try your suggestion later, many thanks!

[mcuneytk]

Dear Shen;
Thanks for the suggestion. Firstly I would like to know if you can continue your run where your run stops or you use your last auto-save data file to continue your run?

Secondly, your suggestion; as much as I know, you also run your case in batch mode and how do you load the case file in batch mode in Linux system ?
Thank you.

Originally Posted by Gang Shen

Hi Ko otto,

Thank for your response!

As for the problem I posted, I have not totally solved it but decreased the interruption frequency to the very low level I can accept. My solution is that when you transfer your case and data files from one operating system(e.g. windows) to a different OS(e.g. Linux) to run the computing, you should load the case files in the new OS, modify the auto-saved directories, save and reopen them. Then you go ahead with your computing. I do not know the reasons but by this method the interruption decreases greatly.

I will also try your suggestion later, many thanks!