My GAMESS geometry optimization did not converge before my hitting my wall time limit. Instead of restarting all the time, I would like to try and string together a series of optimizations together to increase my job throughput.
Orbitals and coordinates from each geometry step during a GAMESS calculation are stored in the .dat file. The most efficient way to string many calculations together is to script the reading of the $VEC and $DATA groups from the .dat file for restarts, and then placing a dependency hold on each job after the first.
I would also recommend knowing how many geometry steps complete each 6 hours or so (depending on your scheduler queue resource limitations), so you can minimize the amount of wall time requested and maximize the number of jobs you have processed through the queue over the course of a week.
Refer to the specific $VEC, $DATA, and method specific restart data (for example, coupled cluster has an option IREST in $CCINP for restart of the CC calculation step) for more information.
As with all questions related to GAMESS, there are method specific keywords and options that may be relevant. Please provide more details about your specific calculation if possible (e.g. chemical system, QM method)