To re-run the previous cycle: see https://ncas-cms.github.io/epoc/model_failures#re-running-the-previous-cycle
To start from an earlier point is more complicated, and involves manually moving files and editing the control nanmelists. You may need to retrieve start dumps from tape. Only the January files are archived.
In this example, assume you want to start from cycle M.
Make sure the atmos restart for cycle M is in place in share/data/History_data.
You also need to make sure the <suite-id>.xhist file is pointing to the correct dump.
If it exists, you can copy this from the cycle before the one you wish to re-run, so M-1.
cd work/<cycle>/coupled/history_archive
cp temp_hist.0001 ~/cylc-run/<suite>/share/data/History_Data/<suite>.xhist
If the dirctory for cycle M-1 no longer exists, then just edit the existing xhist file.
You will also need to set the variable H_STEPIM.
If you have the log file for when the cycle was originally run,
find the timestep for the start of the cycle, e.g.
Atm_Step: Timestep 1866240 Model time: 1986-01-01 00:00:00
You should also check that the ozone files are in place.
Ensure the NEMO restart(s) are in share/data/History_data/NEMOhist.
NEMO will chose the latest file, so make sure any later files are moved aside first.
You also need to have the NEMO control namelist from the previous work directory.
If work/<cycle>/coupled/ does not exist, then create it.
Copy in the NEMO namelist namelist_cfg from a cycle with a 5 min timestep.
Edit the variables nn_it000 and nn_itend.
Again these can be found in the log file for the original job, near the start.
[INFO] Sourcing namelist file from the work directory of the previous cycle
[SUBPROCESS OUTPUT] nn_it000=3746881,
[SUBPROCESS OUTPUT] nn_itend=3755520,
Go to share/data/History_data/CICEhist, and edit ice.restart_file to point to the appropriate file