Opened 8 years ago

Closed 8 years ago

#951 closed error (fixed)

empty output .p* files

Reported by: emb66 Owned by: um_support
Component: HECToR Keywords:
Cc: Platform: <select platform>
UM Version: 7.3

Description

I've recently discovered that some of the .p* output files in my v7.3 runs (performed Aug-Oct) are completely empty. I run two jobs, and the empty files were usually created roughly at the same 'real' time (diffrences of a few minutes):

Jul 30 20:35;
Jul 31 10:35;
Aug 2 12:35;
Aug 4 17:03;
Aug 24 20:08;
Aug 28 20:16;
Aug 31 13:54;
Sep 3 11:13;
Sep 7 11:09;
Sep 8 12:09;
Sep 10 11:12;
Oct 12 17:09;
Oct 16 10:04;
Oct 18 18:49;
Oct 22 15:46 & 22:04;
Oct 29 01:04;

These 'real' times correspond to different model timesteps in the individual jobs and I couldn't find any obvious reason for this behaviour. Would anyone be able to suggest some explanation/help?

Thank you in adavance
Ewa

Change History (5)

comment:1 Changed 8 years ago by grenville

Ewa

Sorry for the late reply - are you still having the problem. If so, please let us know the job name and leave file name.

Grenville

comment:2 Changed 8 years ago by emb66

Dear Grenville,

I've only recently realised that the missing files are produced once the job crashes due to lack of disk space, disturbing the pp file stream, and then the new run just outputs missing pp files right at the beginning of the new run (Surprisingly, the job carries on undisturbed). Eg:

-rw-r—r— 1 emb66 n02 89912634 Oct 21 11:34 xhkbe033.xhkbe.d12295.t013955.leave
-rw-r—r— 1 emb66 n02 98429299 Oct 23 08:54 xhkbe000.xhkbe.d12296.t135145.leave

with the missing files outputted at 22 Oct 22:04, which is the start of the 2nd run. Is there anything I else I can do about it apart from restarting it by setting up a new N-run from a recent start dump?

Thank you
Ewa

comment:3 Changed 8 years ago by grenville

Graham

I see that including the ncas branch didn't help. I took a copy of your job and find that if you move the block of code beginning

LOGICAL
l_subsample

and ending

REAL
min_view_lat

to a position above

NAMELIST/R2SWNCAL/

then no errors are reported in readlsta.f90

Hope this helps.

Grenvile

comment:4 Changed 8 years ago by grenville

Ewa

Apologies - I attached a reply to the wrong ticket!

Grenville

comment:5 Changed 8 years ago by grenville

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

Ewa

Restarting from a recent start dump is about all you can do. It's good to check before hand that you have the disc space needed for the run to complete (this may not be infallible since a disc problem could arise from the n02 quota being exceeded).

Grenville

Note: See TracTickets for help on using tickets.