Main INDEX
Monthly INDEX
PREV
NEXT
Make New Entry,
Make Followup Entry
User name R. Michaels
Log entry time 07:29:51 on December17,2010
Entry number 344788
Followups:
keyword=missing halog end-run problem
Some runs like 9724 and 9727 do not have end-of-run info in halog,
while others like 9722 and 9725 do. Then there are CODA crash
runs like 9723 which obviously don't, but that's understandable.
I investigated the problem. In ~a-onl/epics/runfiles you will see
normally 5 files (ls *9725*), but for the miss-end runs only 4 files,
the missing one being halog_end_*.epics. This file is produced at
the end of the script epicsRunEnd
cat $comfile > $halogfile
cat $epicsfile >> $halogfile
where $comfile is halog_com_*.epics, $epicsfile is End_of_Run_*.epics
and $halogfile is the one that gets missed, namely halog_end_*.epics.
If we look at $epicsfile it's clear what fails:
Consider the lines that retrieve info about DVCS calorimeter, e.g.
log "Anode current of block 200 (c,r:12,8) :" `$CAGET -t -f3 haVMI3128_4_AI9`
You can see that if this reaches the end (block 207) everything
is fine (5 files produced and $halogfile goes to halog.
But if it stops somewhere short (run 9724 stops at block 176
and run 9727 stops at block 190) so the lines that create
$halogfile never happen. Hence no end-run entry in halog.
I'm a little too tired to fix it now, but my speculation is
that the 200+ lines of caget that were added for DVCS calo
take a long time, and if the next run is started quickly
it will "clean up" (via a script designed to kill hanging
processes) the previous instances of the script. The culprit
may be the line "killproc caget emacs" in end_clean. So either
don't do that, or delay that, or improve the logic some other way.
The data are still available but it's not so convenient to miss
the end-run halog entries.