Journaling for data recovery - Lost records.

Ran across an interesting situation at a client on Friday. It would
appear that his UPS was not working and when they took a short power
spike (down/up, not extended) the machine re-IPLed. As part of this
process we lost records that had been written. 
 
We have an invoice, for example, that had items entered, around 7:30 PM,
we have a printed copy of the invoice. We know when the items were
entered because while we don't have the invoice detail record, we do
have the invoice detail detail record. (invoice detail has a child row).
Looking at the messages in QHST we can see that many files were open
when the machine went down, and looking at the last RRN for the invoice
detail file in the message in QHST we can see that the last row in the
file when the machine IPLed was written around 6:45 (rows have
timestamps). The next row by RRN is written around 10PM and the power
failure was 9:21. Based on all this we're quite confident that the
missing invoice detail row was still in memory and hadn't been flushed
to disk and therefore went poof when the power failed. "Poof" being a
technical term. <G>
 
So, here's the question, would journaling have helped? I've used
journaling lots of times are part of commitment control, and I know that
even w/out commitment control I could do an APYJRNCHG to roll forward
changes. However, we didn't discover that rows were missing until the
middle of the following day, so APYJRNCHG wouldn't have helped, we had
thousands of new rows in the file after the point where the missing rows
would have gone. 
 
When the system IPLs I know it uses the SMAPP journal to automatically
recover the access paths, would it do the same with the base tables if
they were journaled? I assume so, but I'm looking for confirmation.
Also, are the entries in the receiver forced to disk when they're
written or might we have lost entries too?
 
On a related topic, what would have happened if we had declared RI
between the invoice detail and invoice detail-detail tables? Would we
now have a child row w/o a parent row?
 
I told the client that he should turn on journaling on all his files as
standard practice, but I got the old "Journaling is too expensive"
response. I know that has changed over the years, but does IBM have any
whitepapers to that effect?
 
-Walden

------------
Walden H Leverich III
President & CEO
Tech Software
(516) 627-3800 x11
WaldenL@xxxxxxxxxxxxxxx <mailto:WaldenL@xxxxxxxxxxxxxxx> 
http://www.TechSoftInc.com <http://www.techsoftinc.com/> 

Quiquid latine dictum sit altum viditur.
(Whatever is said in Latin seems profound.)