| 1 | #!/bin/sh
|
|---|
| 2 | #
|
|---|
| 3 | # ========================================================================
|
|---|
| 4 | #
|
|---|
| 5 | # *
|
|---|
| 6 | # * This file is part of MARS, the MAGIC Analysis and Reconstruction
|
|---|
| 7 | # * Software. It is distributed to you in the hope that it can be a useful
|
|---|
| 8 | # * and timesaving tool in analysing Data of imaging Cerenkov telescopes.
|
|---|
| 9 | # * It is distributed WITHOUT ANY WARRANTY.
|
|---|
| 10 | # *
|
|---|
| 11 | # * Permission to use, copy, modify and distribute this software and its
|
|---|
| 12 | # * documentation for any purpose is hereby granted without fee,
|
|---|
| 13 | # * provided that the above copyright notice appear in all copies and
|
|---|
| 14 | # * that both that copyright notice and this permission notice appear
|
|---|
| 15 | # * in supporting documentation. It is provided "as is" without express
|
|---|
| 16 | # * or implied warranty.
|
|---|
| 17 | # *
|
|---|
| 18 | #
|
|---|
| 19 | #
|
|---|
| 20 | # Author(s): Daniela Dorner 05/2006 <mailto:dorner@astro.uni-wuerzburg.de>
|
|---|
| 21 | #
|
|---|
| 22 | # Copyright: MAGIC Software Development, 2000-2006
|
|---|
| 23 | #
|
|---|
| 24 | #
|
|---|
| 25 | # ========================================================================
|
|---|
| 26 | #
|
|---|
| 27 | # This a script, which launches other scripts (all scripts, that are run
|
|---|
| 28 | # on primary basis
|
|---|
| 29 | #
|
|---|
| 30 |
|
|---|
| 31 |
|
|---|
| 32 | source `dirname $0`/sourcefile
|
|---|
| 33 | printprocesslog "INFO starting $0"
|
|---|
| 34 |
|
|---|
| 35 | set -C
|
|---|
| 36 |
|
|---|
| 37 | echo "" >> $jmscriptlog 2>&1
|
|---|
| 38 | echo "starting jobmanager ("`date`")" >> $jmscriptlog 2>&1
|
|---|
| 39 | makedir $jmlogpath
|
|---|
| 40 | makedir $listpath >> $jmscriptlog 2>&1
|
|---|
| 41 | makedir $lockpath >> $jmscriptlog 2>&1
|
|---|
| 42 |
|
|---|
| 43 | prev=$max
|
|---|
| 44 | user=`whoami`
|
|---|
| 45 | notcount=0
|
|---|
| 46 | while (( $notcount < 100 ))
|
|---|
| 47 | do
|
|---|
| 48 | for (( i=0 ; i < ${#scripts[@]} ; i++ ))
|
|---|
| 49 | do
|
|---|
| 50 | makedir $jmlogpath
|
|---|
| 51 | date >> $jmscriptlog 2>&1
|
|---|
| 52 | echo "sleeping $sleeptime..." >> $jmscriptlog 2>&1
|
|---|
| 53 | sleep $sleeptime
|
|---|
| 54 | dayofweek=`date +%u`
|
|---|
| 55 | hour=`date +%k`
|
|---|
| 56 | source `dirname $0`/sourcefile
|
|---|
| 57 | echo "script: ${scripts[$i]}" >> $jmscriptlog 2>&1
|
|---|
| 58 | # echo " day: $dayofweek hour: $hour" >> $jmscriptlog 2>&1
|
|---|
| 59 | queued=`/usr/local/bin/condor_q -global -format "%s\n" Owner | grep -c $user`
|
|---|
| 60 | queued=`setzero $queued`
|
|---|
| 61 | queuedscript=`/usr/local/bin/condor_q -global | grep $user | grep -c ${scripts[$i]}`
|
|---|
| 62 | queuedscript=`setzero $queuedscript`
|
|---|
| 63 | runningscript=`/usr/local/bin/condor_q -global | grep $user | grep ' R ' | grep -c ${scripts[$i]}`
|
|---|
| 64 | runningscript=`setzero $runningscript`
|
|---|
| 65 | stillinqueue=`echo $queuedscript - $runningscript | bc `
|
|---|
| 66 |
|
|---|
| 67 | totalpno=${pnototal[$hour]}
|
|---|
| 68 |
|
|---|
| 69 | #choose array according to the day of the week
|
|---|
| 70 | case $dayofweek in
|
|---|
| 71 | 0 | 6) pnos=( ${pnoswe[@]} ) ;;
|
|---|
| 72 | *) pnos=( ${pnosweek[@]} ) ;;
|
|---|
| 73 | esac
|
|---|
| 74 | num=`echo "((( $i + 1 ) * 24 ) + ( $hour + 1 ) ) - 24 - 1 " | bc `
|
|---|
| 75 | pnoscript=${pnos[$num]}
|
|---|
| 76 | if [ $prev -eq 0 ]
|
|---|
| 77 | then
|
|---|
| 78 | echo " prev=0 => resetting pnoscript [$pnoscript] to max [$max]" >> $jmscriptlog 2>&1
|
|---|
| 79 | pnoscript=$max
|
|---|
| 80 | fi
|
|---|
| 81 | echo " found $queued jobs in the queue (incl. running jobs) [allowed $totalpno]" >> $jmscriptlog 2>&1
|
|---|
| 82 | echo " found $queuedscript ${scripts[$i]} in the queue (incl. running jobs [$runningscript]) [allowed $pnoscript] - not running: $stillinqueue" >> $jmscriptlog 2>&1
|
|---|
| 83 |
|
|---|
| 84 | if [ "$queued" -gt "$totalpno" ]
|
|---|
| 85 | then
|
|---|
| 86 | cont >> $jmscriptlog 2>&1
|
|---|
| 87 | else
|
|---|
| 88 | if [ "$queuedscript" -gt "$pnoscript" ]
|
|---|
| 89 | then
|
|---|
| 90 | cont >> $jmscriptlog 2>&1
|
|---|
| 91 | fi
|
|---|
| 92 | fi
|
|---|
| 93 |
|
|---|
| 94 | # check if there's something to do
|
|---|
| 95 | getstatus >> $jmscriptlog 2>&1
|
|---|
| 96 | echo " $numproc ${scripts[$i]} still do to" >> $jmscriptlog 2>&1
|
|---|
| 97 | if [ "$numproc" = "" ]
|
|---|
| 98 | then
|
|---|
| 99 | prev=0
|
|---|
| 100 | cont >> $jmscriptlog 2>&1
|
|---|
| 101 | fi
|
|---|
| 102 | prev=$max
|
|---|
| 103 | if [ $numproc -lt $stillinqueue ]
|
|---|
| 104 | then
|
|---|
| 105 | echo " numproc($numproc) -lt stillinqueue($stillinqueue)" >> $jmscriptlog 2>&1
|
|---|
| 106 | cont >> $jmscriptlog 2>&1
|
|---|
| 107 | fi
|
|---|
| 108 |
|
|---|
| 109 | date=`date +%Y-%m-%d`
|
|---|
| 110 | echo " committing 1 ${scripts[$i]} to condor" >> $jmscriptlog 2>&1
|
|---|
| 111 | if ! /usr/local/bin/condor_submit -a path=$scriptspath -a prog=${scripts[$i]} -a date=$date -a dir=$runlogpath $scriptspath/run.condor 2>> $jmerrorlog
|
|---|
| 112 | then
|
|---|
| 113 | date >> $jmerrorlog
|
|---|
| 114 | nail -s 'condor problem' $erradrs < $jmerrorlog
|
|---|
| 115 | echo "condor is not working -> sleeping $errorsleeptime" >> $jmscriptlog 2>&1
|
|---|
| 116 | printprocesslog "ERROR submitting ${scripts[$i]} to condor failed"
|
|---|
| 117 | sleep $errorsleeptime
|
|---|
| 118 | fi
|
|---|
| 119 | date >> $jmscriptlog 2>&1
|
|---|
| 120 | echo "" >> $jmscriptlog 2>&1
|
|---|
| 121 | done
|
|---|
| 122 | done
|
|---|
| 123 |
|
|---|
