Hi I am running two suites u-cn515 and u-cn440 which are both failing at the pptransfer task and I cannot see why.
The files appear to have been transferred. The job.out script says the transfer was successful. There is no obvious error in the job.err script.
/home/n02/n02/schn02/cylc-run/u-cn440/log/job/18411001T0000Z/pptransfer/job.out
/home/n02/n02/schn02/cylc-run/u-cn440/log/job/18411001T0000Z/pptransfer/job.err
Can you advise as to what the problem is?
Thanks
Andrew
Hi Andrew,
/work/n02/n02/schn02/cylc-run/u-cn440/log/job/18411001T0000Z/pptransfer/10/job.status indicates the transfer has now succeeded. Prior to that it has failed due to connection issues which I guess would be due to the JASMIN maintenance that has been going on the last few days.
/work/n02/n02/schn02/cylc-run/u-cn515/log/job/17841001T0000Z/pptransfer/05/job.status has also succeeded.
The reason it is indicating a failure even though the task is succeeding is due to running the pptransfer on the login nodes & how cylc task polling works.
For cylc to be able to correctly poll the running pptransfer task it has to login to the same login node that the task is running on, this means you can’t using login.archer2.ac.uk for the pptransfer tasks as this picks at random one of the 4 ARCHER2 login nodes so you’re not guaranteed to land on the same node each time.
In the archer2.rc file add the following to the [[PPTRANSFER_RESOURCE]] family to force the pptransfer task to run on a specific login node.
[[[remote]]]
host = login3.archer2.ac.uk
You can pick any login node numbered 1 to 4.
Cheers,
Ros.