YellowSheet: mrenna

File mrenna, 1.6 KB (added by anonymous, 6 years ago)
Line 
1The change from the "old" fermilab batch system
2to condor was pretty straightforward, so I have
3about 4 years experience now!
4
5The basic idea is that almost everything has to
6be done remotely, but your beautiful code is
7set up to do this nicely.
8
9You package everything up you need for each run --
10which is just a few files.  Here is a typical condor
11job description file:
12
13universe = vanilla
14Notification = Error
15executable =
16/farm/run2mc_stage01/condor/W2/SubProcesses/Pug_e+vedg/ajob8.sh
17environment = F77=g77
18transfer_output = true
19transfer_error = true
20transfer_executable = true
21should_transfer_files = YES
22transfer_input_files =
23ajob8,input_app.txt,madevent,input-card.dat,cteq6l1.tbl,randinit
24WhenToTransferOutput = ON_EXIT_OR_EVICT
25input = /dev/null
26output = $(Cluster).$(Process).out
27error =  $(Cluster).$(Process).err
28log =    $(Cluster).$(Process).log
29queue
30
31After each step (survey, refine, refine), a bunch of files come
32back into the appropriate directories.  THEN, I run a
33shell script locally:
34
35#!/usr/local/bin/bash
36
37./bin/march
38cd SubProcesses
39../bin/sumall
40../bin/combine_events
41cd ..
42
43
44"march" is a script I wrote that "marches" through all of the
45directories, untars alot of tarballs, and basically prepares the
46output for combine_events.
47
48Also, I found that it was more efficient to deal with
49"gzipped" events.dat files.  There is actually a library
50out there that lets you READ gzipped files in fortran.
51
52As far as I can tell, there is no GOOD way to avoid the
53local processing step between survey, refine, refine.
54
55People won't like that, but that is because they haven't
56really thought about the problem.
57