Thread: question about PANIC: hash table "Shared Buffer Lookup Table" corrupted

hi

I use Postgresql 7.4.19 wiht 8G memory on Redhat5 x86_64.
And I modify the postgresql.conf file,
when I set
shared_buffers = 786432                # min 16, at least
max_connections*2, 8KB each
(6G)
and import data.
[ PANIC:  hash table "Shared Buffer Lookup Table" corrupted ]
occur.

I set the shared_buffers to
shared_buffers = 393216  (3G)
no error occur.

Thanks for any idea.

"bh yuan" <bhyuan@gmail.com> writes:
> I use Postgresql 7.4.19 wiht 8G memory on Redhat5 x86_64.
> And I modify the postgresql.conf file,
> when I set
> shared_buffers = 786432                # min 16, at least

7.4 does not support more than 2Gb (or maybe 4Gb, I forget) of shared
memory.  Furthermore, it so little imagines that you might want that
much that it doesn't even guard itself against overflow.  This is not
really a problem because its buffer management algorithms are not good
enough to support more than a few tens of thousands of shared buffers
anyway.

In short: don't do that.

Modern versions of PG (say, 8.1 or later) are somewhat competent with
large amounts of shared buffers.

Why are you even trying such an obsolete version of PG on RHEL5?
What Red Hat ships in RHEL5 is 8.1, and even that is pretty old in
the view of most people on this list.

            regards, tom lane