SMKINVEN Killed in ptfire 2017

SMOKE Run Errors and Warnings
1

Hello all,

I am running ptfire emission for January 2017 in Central Florida region. But it seems that SMKINVEN is getting killed due to high memory requirements. A probable reason could be that the inventory file (ptday) is quite big. I am attaching my run scripts, inventory file list that I used,
and error message here. I will be grateful if anyone can help me resolve this issue.

Hasibul
Error.txt (143.1 KB)
smk_point_nctox.csh (5.4 KB)
smkinven.ptfire.nctox.txt (25.0 KB)
ASSIGNS.nctox.cmaq.cb05_soa.us12-nc.txt (25.6 KB)
ptday.point.txt (150 Bytes)
ptinv.point.txt (156 Bytes)

2

HI,
Were you able to get this working?

If not, I pulled the following information out of the smkinven.ptfire.nctox.txt log file that you provided in case others can provide assistance.
Successful open for emissions file:
/lustre/fs0/home/mhasan/Parallel/cmaq/Build_CMAQ/LIBRARIES/data/inventory/nei2005/point/ptday_ptfire_2017NEI_20200206_CONUS_CAPs_07apr2020_nf_v1
Value for HOURLY_TO_DAILY not defined;returning default: FALSE
Value for HOURLY_TO_PROFILE not defined;returning default: FALSE

 *** ERROR ABORT in subroutine GENPDOUT:CHECKMEM
 Failure allocating memory for "CODEA":  STATUS=        41

You had additional information in the Error.txt file:
segfault at 98 ip 00002ae54b18fbd6 sp 00002ae54b1f2bb0 error 4 in ld-2.17.so[2ae54b184000+22000]
[2210197.139279] Call Trace:
[2210197.139295] [] dump_stack+0x19/0x1b
[2210197.139300] [] dump_header+0x90/0x229
[2210197.139309] [] ? default_wake_function+0x12/0x20
[2210197.139316] [] ? pollwake+0x73/0x90
[2210197.139325] [] oom_kill_process+0x25e/0x3f0
[2210197.139332] [] ? cpuset_mems_allowed_intersects+0x21/0x30
[2210197.139341] [] mem_cgroup_oom_synchronize+0x546/0x570
[2210197.139347] [] ? mem_cgroup_charge_common+0xc0/0xc0
[2210197.139353] [] pagefault_out_of_memory+0x14/0x90
[2210197.139357] [] mm_fault_error+0x6a/0x157
[2210197.139364] [] __do_page_fault+0x491/0x500
[2210197.139369] [] do_page_fault+0x35/0x90
[2210197.139376] [] page_fault+0x28/0x30

3

I still having difficulty with this issue. I am running for the CONUS12 domain for January 2017. Here are the log files, assign files and run files:
ASSIGNS.nctox.cmaq.cb05_soa.us12-nc.txt (25.6 KB)
slurm-4281397.txt (20.8 KB)
smk_point_nctox.txt (5.2 KB)
smkinven.ptfire.nctox.txt (41.7 KB)

I will be grateful if anyone can help me resolve this issue.

Hasibul

4

The NC Tox case is very old and is no longer supported.

A full set of 2017 scrips and data files are available here:

https://gaftp.epa.gov/Air/emismod/2017/

You should not have any NCTOX files if you use this case.

5

In addition to switching to the newer 2017 platform scripts, we note that we normally do not process the annual and daily fire inventories in a single instance of Smkinven. In our run scripts, we run Smkinven to process the annual FF10 first, and then once (per month) to process the daily FF10.

Since you are doing both together, you are probably running out of memory and that is why the program is killed.

6

Even if the file name is nctox, I am using 2017 data. Here are the ptinv and ptday files:

ptday.point.txt (596 Bytes)
ptinv.point.txt (302 Bytes)

Hasibul

7

How can I change the script to process only January?

Hasibul

8

Most of these scripts have a months setting, so you would change that variable to run only month 1. Write back if you can’t find that setting…