During a recent
system upgrade, the system manager purged a disk
that contained our Run Unit Journal (RUJ)
files. Unfortunately, this resulted in some corruption in our
database that was not immediately discovered. Worse yet, all of our
backup files now contain corrupted images of the database. Is there
any way to recover from this situation?
A:
The
problems that you have experienced indicate that there are several
weaknesses in your environment:
The
frequency with which DBO/VERIFY
was being executed.
The
backup and retention cycle for your database files.
Lack
of adequate controls over access to and retention of your RUJ
files.
However,
all of that is outside of the scope of your immediate problem,
which is fixing your database. The DBO/VERIFY utility
indicates that your database contains corruption of several index
nodes. While we can
theoretically patch the database using the DBO/ALTER (DBMS low-level
patching utility), the extent of the corruption will make this a
slow and error-prone task.
To
solve this problem, we used a combination of tools that allowed us
to automate much of our work.
We
wrote a special purpose program to obtain the pointer clusters
from the current records, and store them in a RMS file.
Using
a special version of SCI's DBInitialize program, we
were able to remove the pointer clusters and index nodes of the
affected set types.
We
wrote a special purpose program that read the pointer
information contained in step 1 to reconnect the pointer
clusters of the appropriate sets.
