Issues Restarting CESM from Past Restart Files with GLC Component Errors

[jentus]

Dear all：
I have been running CESM using the startup mode and continually submitting jobs to generate output and restart files for advancing the simulation. Currently, I have a sequence of restart files from various time points. If I attempt to continue the simulation from the latest restart file (e.g., 0002-11-11), it works without any issues. However, when I try to restart the simulation from a past restart file (e.g., 0002-10-02), the model fails to run.

The issue appears to be specific to the GLC component, which fails to advance while other components successfully progress. This desynchronization causes the model to crash. I have attempted to restart using both the startup and branch modes, but the same issue persists when using past restart files.

My question is: How can I restart CESM from past restart files successfully? Is this a known issue with the GLC component, and is there a way to address it?

Thank you for any advice or suggestions!

 

[dbailey]

Are you copying all of the files from the rest/0002-10-02 directory including the pointer files into your run directory?

 

[jentus]

Are you copying all of the files from the rest/0002-10-02 directory including the pointer files into your run directory?

Thank you, dbailey！ I re-copied all the previous restart files and re-ran the case using the "branch" mode, and it worked successfully. However, one thing still confuses me. In the cases where my runs failed previously, the model would delete some of the files from the previous runs (even though the new file names were different from the old ones). I don't quite understand why this happens.

 

[dbailey]

I guess I was confused about whether you were doing a branch or a hybrid. Usually we make use of RUN_REFCASE and RUN_REFDATE in env_run.xml to point to the run we are branching / starting from. For a hybrid run, this requires less restart information, whereas a 'branch' is an exact continuation of the old run.

 

[slevis]

one thing still confuses me. In the cases where my runs failed previously, the model would delete some of the files from the previous runs (even though the new file names were different from the old ones). I don't quite understand why this happens.

If I understand your concern correctly, I think that the files from earlier stages of the simulation that seem to disappear, should now appear in the /archive directory. From the case's /run directory, type "cd ../../archive"