Recovery at a DR Site

Discussion created by ca.portal.admin on Feb 17, 2005
Latest reply on Feb 28, 2005 by ca.portal.admin
Hi All,
Sorry for the long post, but I need to explain the situation before asking
the question.
We are now doing data replication for disaster recovery, so I no longer
need to rely on journals to foll forward at a DR site. For those who are
not familiar with this facility, it simply means there is a remote site with
duplicate DASD which is constantly kept in sync with the main data center.
During the current DR test, I noticed that once the IDMS regions
WARMSTARTed, one region had a full journal. This could only mean that a
journal offload was running at the main site when the decision was made to
""snapshot"" the data (all IDMS regions are active at the time). I simply ran
an offload at the DR site and all was fine.
However, this got me to thinking. What if the journal offload was
performing the condense of the journal file when the ""disaster"" occurred?
This would mean that the journal file at the DR site was in an unknown state
- having been constantly copied while the condense was occuring. If the
region (at DR) needed data within that journal file in order to restart (a
very real possibility), the restart could very well fail.
So, I'm thinking a simple solution to this problem would be to modify the
journal offload job. It would copy the journals before the offload step,
and then delete the copy once the offload was successful. In this way, if a
disaster occurred, I could check for the presence of a backup journal at the
DR site before bringing up IDMS. If it is there, that would indicate that
an offload was occuring at the time of the disaster and I would copy the
backup journal file (only the FULL one) over the top of the real journal
file, then start the regions.
One question which comes to mind is whether I back up all the journals
before the offload or just the one which is full. If I only wanted to back
up the journal which would get offloaded, I'm guessing I would need to write
a program to determine the full journal (with the lowest sequence number)
and somehow generate the correct control cards for the backup.
If I decide to back up all the journal files before the offload, I would
then need to determine which journal should get overlayed at the DR site
(ie. which journal was being offloaded at the main site). This could be
done by running a simulated offload against the backup journals at the DR
site and see which one would get chosen.
I think backing up all the journals is the easiest method. Does anyone
see anything wrong in this approach?

Joe Lupico
IDMS System Support
""Our World is a Happy World""

IDMS Public Discussion Forum


Re: Recovery at a DR Site
" For this particular client, it is real time, with a ""delay"" of
approximatly 40 seconds.

Joe Lupico
IDMS System Support
""Our World is a Happy World""