[gmx-users] Restarting a crashed MD run

[Mark Abraham]

bmmothan at ucalgary.ca wrote:
> sorry to ask again, I hope this is my last question about this.
> 
> since I know now that the restarted run I did started from 0 time. I
> killed the job and it wrote velocity for the last frame. So since i knew i
> have to contiune this run that I just killed, I used tpbconj and it wrote
> a new tpr file and it said it will restart it from the time I killed the
> job ( half the simulation). SO i think I am set now to continue the
> simulation this way.

The first time you used it, it will also have told you the frame it was 
going to be able to start from. The output messages are usually very 
thorough and helpful. Read them carefully and think about what they 
mean! :-)

> Thank you. I will save some energies and velocities in future simulations.
> 
> but in this email, my job was not crashed..I terminated it and it wrote
> the last velocities and when i wanted to contiune it using pbdconj, it
> said it will finish the simulation from the time I terminated it.
> 
> so is this accurate enough to finish it?

Maybe. To get an exact continuation of some runs with temperature and/or 
pressure coupling you need to supply an .edr file. If you don't give one 
and it wanted one, tpbconv will warn you.

Had you read the man page to tpbconv closely before using it, you would 
have been aware of these issues already. It describes why you had all of 
the problems that you had. You would have saved yourself and the mailing 
list some time :-) Please make a habit of searching manuals, man pages 
and mailing list archives before asking.

Mark