Hello all,
Background:
DB resides on a mirrored array, HD1 goes bad and before it can be
replaced, HD2 starts throwing hardware errors. We were able to get a
DD from HD2 and have restored the box quite successfully with the
exception of the pg db. There WAS corruption to the filesystem. Since
the last backup, a significant number of records have been added to
the db. If there is a way to recover some data, it would likely be
easier to reconcile and fix manually than reenter. Box is FreeBSD 8,
postgresql version 8.1.
Progress so far:
On new hardware, installed pg8.1 and restored data directory and
attempted to start pg. After fixing some config differences, it
complained of a missing pg_clog/xxx file. After searching the archives
for hints, added a file of zero's and pg starts successfully. The db
in question exists with what appears to be intact structure. When
attempting to dump the db ran into the error: "could not open relation
1663/16385/16400: No such file or directory" and run into similar
errors (with different relation numbers) trying to select * from
certain tables.
I'm stuck here. Assuming the damage to filesystem, what is the best a
way forward? Does the relation error indicate damage to the data? or
to an index??? Is there a process I can do to try and salvage partial
data?
Thanks for any help/tips/suggestions.
Roland Wells
