"Chad Thompson" <chad@weblinkservices.com> writes:
> It seems that my postgres server was rebooted in the middle of a vacuum full.
> DEBUG: server process (pid 22196) was terminated by signal 11
You need to isolate what's causing that repeatable backend crash.
(A first guess is that some page of data got corrupted during the system
crash --- was this due to a power failure, or just some cowboy hitting
ctrl-alt-del? But we shouldn't leap to conclusions.)
The standard full-up procedure for investigating backend crashes of
unknown origin is to get a stack backtrace and post it for inspection on
pgsql-hackers or pgsql-bugs. In this case, though, since we suspect
corrupt data a reasonable shortcut is to try "vacuum verbose analyze"
and see if it makes it all the way through or not. If not, the table it
was working on when it died is the (or at least a) locus of corruption.
If you're really lucky, the corruption is in an index or a table you can
rebuild easily. If not, well, we can probably help you scavenge what's
recoverable ... but it's getting a bit far afield for pgsql-novice.
Please do the preliminary investigation and if you need more help,
report in on pgsql-hackers.
regards, tom lane
