Main INDEX
Monthly INDEX
PREV
NEXT
User name R. Michaels
Log entry time 00:30:46 on August25,2003
Entry number 113353
This entry is a followup to: 113193
Followups:
keyword=missing runs due to data9 crash
I investigated /adaql2/data9 to see when it died and what the
implications were. It seems to have died sometime between 14:00
and 21:30 on Aug 17. There are three runs potentially lost:
3683, 3692, and 3701. But I don't know what is wrong with that
disk, so they may not be lost (we'll have to attempt surgery after
the experiment). After 3701, every 9 runs were a "CODA crash"
since data9 was not available and prestart would fail. The first
such run is 3709 (which is not skip-by-9 because 3706 was a file
splitted run).
This problem raises the question of why not make our data disks
RAID arrays, if possible, to reduce the chance of losing data ?
There were historical reasons why not, but they may not be relevant now.
Also, maybe we should copy data to MSS immediately instead of waiting
for a time. I'll investigate these possibilities and try to recover
the missing runs during the break.