Calculation Restart Problems

[attene]

Hello,

I tried a restart and It may worked but one thing it is not clear: my residual.csv shows the same number of iteration as the first calculation; although the file was written at the end of the restart calcultion.

I did two steady simulation:
1) 10 iteration
2-restart) 15 iterations

Regards,
FA

[Luciano Garelli]

Hello,

It's strange... you should have the number of iterations of each run, like in the following example.

You can force the flush of the listing, residuals ant time_stat using the control_file, check the user guide there is info about the use of the control_file.

Regards,

Luciano

[attene]

Hi Luciano,

thanks you for your answer!
I was thinking I might have made a mistake launching the restart: maybe I should have launched it from the checkpoint directory!? I will try doing that and looking at the use of the control file.

Regards,

FA

[attene]

Hi Luciano,

I read about the use of the control file in the user guide. All the main options are listed but I do not have a clear idea of what suits my issue. What does the flush option do?

Regards,

FA

[attene]

Hi everybody,

problem solved! I was launching the job from the wrong directory.
The restart was not pointing at the checkpoint directory of the previous calculation.

Regards,

FA

[attene]

Hi everybody,

still a question regarding a restart from a job (RESU/YYYYMMdd-hhmm) which was restarted; please forgive the play of words: in practical terms in the folder I am considering (YYYYMMdd-hhmm) there are both a checkpoint and a restart directories.
If I want to do a new restart from this job; do I have to point the new job to the restart directory or the checkpoint directory?

Both checkpoint and restart directories contain the files for a restart (main and auxiliary).

Hope I have been clear enough...

Best regards,

FA

[attene]

Hi everybody,

still a question regarding a restart from a job (RESU/YYYYMMdd-hhmm) which was restarted; please forgive the play of words: in practical terms in the folder I am considering (YYYYMMdd-hhmm) there are both a checkpoint and a restart directories.
If I want to do a new restart from this job; do I have to point the new job to the restart directory or the checkpoint directory?

Both checkpoint and restart directories contain the files for a restart (main and auxiliary).

Hope I have been clear enough...

Best regards,

FA

Regarding my post I think the answer is the checkpoint directory because the restart is created in YYYYMMdd-hhmm when a job "is called" from the checkpoint directory.

Best
FA

[Yvan Fournier]

Hello,

A computation reads a restart (if present) and generates a checkpoint.

The restart in an execution directory is usually a symbolic link to a checkpoint from a previous run (except maybe on Windows where it might be a copy).

So you probably want the checkpoint.

Regards,

Yvan