I have been trying to run the job u-cz105, but it keeps getting ‘stopped with ‘running’’ and I think this is due to my disk quote on Puma. I have 12G, but running the job builds up the cylc-run directory very quickly and after 5 years it is taking up 5G on Puma alone. Is there a way to clean this directory as I run/ are there bits I can delete and not affect the run?
You have Extra Diagnostic Messages switched on so that is making the atmos_main log files rather large. We only suggest switching this on for diagnosing problems, otherwise change it to Minimum or Normal output messages & warnings.
You can also delete old log files for cycles that have completed on PUMA under the cylc-run/u-cz105 directory. If you really do want to keep extra diagnostic messages on then I would suggest telling cylc not to pull back the job logs from ARCHER2 by adding in the
[[HPC]] section of the
retrieve job logs = False
Thank you! I had switched that on previously to find an error and must have missed switching it back off again.
When deleting old log files, can I just check this is the files within log.20230828T170855Z labelled job-19900101T0000Z.tar.gz for example?
Or is it the log files in cylc-run/u-cz105/log/job/19940101T0000Z for example?
Yes you can delete the ones labelled
Housekeeping tars/gzips up the
cylc-run/u-cz105/log/job/19940101T0000Z regularly, I’m not sure if it will complain if the directory isn’t there or not without trying.