Merge pull request #9 from ACME-Climate/worleyph/Machines/mira-mkbatc…

…h-fix Fixing bug that can cause job scripts on Mira to exit before finishing postprocessing
E3SM-Project · Sep 18, 2014 · ed554ed · ed554ed
2 parents 5342f0c + c4c15fe
commit ed554ed
Show file tree

Hide file tree

Showing 2 changed files with 17 additions and 8 deletions.
diff --git a/scripts/ccsm_utils/Machines/mkbatch.mira b/scripts/ccsm_utils/Machines/mkbatch.mira
@@ -192,7 +192,11 @@ runjob --label short -p \${procs} -n \${ntasks} \${LOCARGS} --envs BG_THREADLAYO
 else
                                     \$EXEROOT/cesm.exe >&! cesm.log.\$LID
 endif
-if (\$syslog_id != 0) kill \$syslog_id
+if (\$syslog_id != 0) then
+  if { kill -0 \$syslog_id } then 
+    kill \$syslog_id
+  endif
+endif
 
 echo "\`date\` -- CSM EXECUTION HAS FINISHED" 
 set sdate = \`date +"%Y-%m-%d %H:%M:%S"\`

diff --git a/scripts/ccsm_utils/Machines/syslog.mira b/scripts/ccsm_utils/Machines/syslog.mira
@@ -18,10 +18,12 @@ while ($outlth < 1)
   set outlth = `wc \-l $run/cesm.log.$lid | sed 's/ *\([0-9]*\) *.*/\1/' `
 end
 
-set remaining = 0
-set rem_hours = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\1/' `
-set rem_mins = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\2/' `
-set rem_secs = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\3/' `
+set rem_hours = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\1/' `
+set rem_mins = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\2/' `
+set rem_secs = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\3/' `
+if ("X$rem_hours" == "X") set rem_hours = 0
+if ("X$rem_mins" == "X")  set rem_mins = 0
+if ("X$rem_secs" == "X")  set rem_secs = 0
 @ remaining = 3600 * $rem_hours + 60 * $rem_mins + $rem_secs
 cat > $run/Walltime.Remaining <<EOF1
 $remaining $sample_interval
@@ -38,9 +40,12 @@ while (1)
   cp -p -u $timing/* $dir
   chmod a+r $dir/*
   sleep $sample_interval
-  set rem_hours = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\1/' `
-  set rem_mins = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\2/' `
-  set rem_secs = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *\([0-9]*\):*\([0-9]*\):*\([0-9]*\) */\3/' `
+  set rem_hours = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\1/' `
+  set rem_mins = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\2/' `
+  set rem_secs = `qstat -lf $jid | grep TimeRemaining | sed 's/ *TimeRemaining *: *0*\([0-9]*\):*0*\([0-9]*\):*0*\([0-9]*\) */\3/' `
+  if ("X$rem_hours" == "X") set rem_hours = 0
+  if ("X$rem_mins" == "X")  set rem_mins = 0
+  if ("X$rem_secs" == "X")  set rem_secs = 0
   @ remaining = 3600 * $rem_hours + 60 * $rem_mins + $rem_secs
   cat > $run/Walltime.Remaining << EOF2
 $remaining $sample_interval