You can see my initial input file, submit file and restrart files attached. Whatever I try to restart my disturbed jobs, I get this error with following input files:
FileIO operation on non-existent file.
FileIO: IOper= 2 IFilNo(1)=-20502 Len= 100 IPos= 4 Q= 244815552
Error termination in NtrErr:
NtrErr Called from FileIO.
Initial input file:
%mem=15GB
%NProcShared=40
%RWF=/scratch/orhan/Rh_Raman_20
%NoSave
%chk=checkpoint.chk
#p opt pbepbe/lanl2dz scf=qc
restart input file 1:
%mem=15GB
%NProcShared=40
%RWF=/scratch/orhan/Rh_Raman_21
%NoSave
%chk=checkpoint.chk
#p opt=restart pbepbe/lanl2dz scf=qc
restart input file 2:
%mem=15GB
%NProcShared=40
%RWF=/scratch/orhan/Rh_Raman_23
%NoSave
%chk=checkpoint.chk
#p opt Geom=AllCheck Guess=Read pbepbe/lanl2dz scf=qc
Such an error in many cases means that the Gaussian cannot read the file. The reason for this may be: 1. The file does not exist 2. The file is corrupted 3. Gaussian versions do not match. The last point is of course unlikely, the first one is easy to check, and the chkchk utility will help you with the second one.
Dear Orhan Kıbrıslı
If you are running the job in a cluster, probably you are doing the calculation in a scratch directory. In this case, check if the script is copying the .chk file to the work directory or if that directory is common to the computing nodes (maybe they are not able to read the file)
However, if your system was able to perfom some optimization steps, I would take the last geometry and use it as starting geometry for a new optimization. In this way, you avoid this kind of problems. Also, you can check if the last geometry is reasonable or not.
Best,
Fernando
include the keyword "Opt=Restart". This keyword tells Gaussian to restart the optimization job using the information in the output file from the previous job. This may can work.
Bharat Prasad Sharma keyword "Opt=Restart" tells Gaussian to restart the optimization job using the information in the chk file, but as you can see from the original post, the program can't find it.
Alexey N. Masliy the file is in the folder that I run the job but I don't know if it is corrupted or not but I tried many times so at least one of them should have been working. what is chkchk utility?
Fernando Aguilar-Galindo yes, I am running it on a cluster. I was using project folder instead of scratch since the output is not that large. I will try to run it on scratch. but script creates the checkpoint file in my working directory I don't understand why the software cannot read it in there. should I delete %chk=checkpoint.chk line in the restart file? but how am I supposed to get the new checkpoint file then?
Orhan Kıbrıslı chkchk is a standard Gaussian utility that checks the integrity and contents of checkpoint files. https://gaussian.com/chkchk/
Since your calculations are carried out on clusters, the problem may be as follows. I don't know how the calculations are organized specifically for you, but usually the task file (.gjf) is copied to the cluster, the calculations are carried out jn cluster's disk and then the results are returned to user and, apparently, only the gjf file is copied and chk is not copied, so Gaussian can't locate it when the task is solved on the cluster. Contact your system administrator to find out how to solve this problem. You may need to specify the full path to the chk file.
You can, as well, just find the best optimised geometry from your output file and reinput this as your starting geometry in a new gaussian input file.
I ran a lot of gaussian jobs on clusters, and commonly came across the kind of errors you're seeing. It's difficult to comment on in detail as we don't know your cluster's environment; and to be honest, the above is just as easy as anything else. Also, opt=restart uses the last completed cycle's geometry, which is not necessarily always the best.
I rather suspect one of your restarts has messed up the checkpoint file, which as someone else mentioned can be looked at with the chkchk utility.
- Badges
- Science topic
- Similar topics
- Filing