Re: FATAL: cache lookup failed for access method 6881280

Brock Williams <brock(at)cotcomsol(dot)com> writes:
> Tried to post directy to the list but I got a bounce back. Must be too
> big...
> Here it is posted to my web site:
> http://www.cotcomsol.com/~brock/postgresql_debug.txt

Hmm, this is quite odd. You have

Item 22 -- Length: 156 Offset: 4760 (0x1298) Flags: USED
XMIN: 1 CMIN: 0 XMAX: 0 CMAX|XVAC: 0
Block Id: 1 linp Index: 22 Attributes: 24 Size: 32
infomask: 0x0910 (HASOID|XMIN_COMMITTED|XMAX_INVALID)

1298: 01000000 00000000 00000000 00000000 ................
12a8: 00000100 16001800 10092000 4f410000 .......... .OA..
12b8: 70675f70 726f635f 70726f6e 616d655f pg_proc_proname_
12c8: 61726773 5f6e7370 5f696e64 65780000 args_nsp_index..
12d8: 00000000 00000000 00000000 00000000 ................
12e8: 00000000 00000000 00000000 00000000 ................
12f8: 0b000000 00000000 01000000 00006900 ..............i.
--------
1308: 01000000 00000000 94000000 00c0da44 ...............D
--------
1318: 01000000 00000000 00006900 04000000 ..........i.....
--------
1328: 00000000 00000000 00000000 ............

A non-broken 8.0 installation has

Item 22 -- Length: 156 Offset: 4760 (0x1298) Flags: USED
XMIN: 1 CMIN: 0 XMAX: 0 CMAX|XVAC: 0
Block Id: 1 linp Index: 22 Attributes: 24 Size: 32
infomask: 0x0910 (HASOID|XMIN_COMMITTED|XMAX_INVALID)

1298: 01000000 00000000 00000000 00000000 ................
12a8: 00000100 16001800 10092000 4f410000 .......... .OA..
12b8: 70675f70 726f635f 70726f6e 616d655f pg_proc_proname_
12c8: 61726773 5f6e7370 5f696e64 65780000 args_nsp_index..
12d8: 00000000 00000000 00000000 00000000 ................
12e8: 00000000 00000000 00000000 00000000 ................
12f8: 0b000000 00000000 01000000 93010000 ................
--------
1308: 4f410000 00000000 94000000 0060da44 OA...........`.D
--------
1318: 00000000 00000000 00006900 04000000 ..........i.....
--------
1328: 00000000 00000000 00000000 ............

where I underlined the parts that are different. The first different
word is the relam field, which leads directly to the reported error
message, and the other two changes are bad news as well. (Note: the
difference at offset 1314 is the reltuples field, which could
legitimately be different across installations, and your value matches
the other pg_proc indexes so it's presumably OK.)

I have no idea what happened here ... but if you can put those values
back the way they were you'll probably be all right. You'll need to
shut down the postmaster before editing the file (else it may have the
page cached in shared memory), and if there is a pg_internal.init file
in the same directory delete it.

regards, tom lane