Opened 4 years ago

Closed 4 years ago

#1585 closed help (answered)

OOM Error on Archer Running UM8.4+UKCA

Reported by: pliojop Owned by: um_support
Component: UKCA Keywords: UM, UKCA 8.4
Cc: Platform: ARCHER
UM Version: 8.4

Description

Morning,

Since late March I have been running teh UM at version 8.4 with UKCA based on Luke Abraham's release job.

I had been initially limited to running the job in 8 hour chunks, getting 3 months & 10 days in those periods. On Friday I switched to 24 hour length jobs, as Luke had informed me that the issue that had prevented me running for more than 8 hours had been resolved.

However, both my jobs experienced a couple of OOM errors over the weekend causing them to crash out after between 13 and 15 hours.

I have resubmitted the jobs again this morning as 8 hour length jobs, and will add to the ticket if they experience an issue.

Thanks

James

Change History (3)

comment:1 Changed 4 years ago by grenville

James

This sounds like a problem we have seen with the Ceray cce8.2.1 complier - please tell us the jobid.

Grenville

comment:2 Changed 4 years ago by pliojop

Hi Grenville,

Sorry, xlaya and xlayb were the job IDs!

James

comment:3 Changed 4 years ago by grenville

  • Resolution set to answered
  • Status changed from new to closed

James

If you rebuild the entire model now, you will pick up cce8.3.7 as the compiler - you won't need any special hand edits, however, you will need to ensure that the default cray-netcdf module is loaded, rather than cray-netdf/4.3.1.

If your model is running OK, there is no need to rebuild 'though.

Grenville

Note: See TracTickets for help on using tickets.