Re: invalid page header in pg_statistic

How do I fix this 7.4.6 issue short of initdb?

invalid page header in block 110 of relation "pg_statistic"

I looked at the block via pg_filedump (included below), and it
does not appear to me to be corrupted, so not sure what I would
zero out, if anything.

TIA.
Ed

*******************************************************************
* PostgreSQL File/Block Formatted Dump Utility - Version 3.0
*
* File: 16408.backup-20070207-01:11:58PST
* Options used: -y -f -R 110 110
*
* Dump created on: Wed Feb 7 01:56:04 2007
*******************************************************************

Block 110
********************************************************
<Header> -----
Block Offset: 0x000dc000 Offsets: Lower 160 (0x00a0)
Block: Size 8192 Version 1 Upper 8192 (0x2000)
LSN: logid 178 recoff 0xf577b4e8 Special 8192 (0x2000)
Items: 35 Free Space: 8032
Length (including item array): 164

0000: 000000b2 f577b4e8 00000028 00a02000 .....w.....(.. .
0010: 20002001 3da00000 3ae00000 37300000 . .=...:...70..
0020: 34d00000 32100000 2e600000 2c000000 4...2....`..,...
0030: 29400000 25900000 23300000 20700000 )@..%...#0.. p..
0040: 1cc00000 1a600000 17a00000 13f00000 .....`..........
0050: 11900000 0ed00000 0b200000 08c00000 ......... ......
0060: 06000000 02500000 1bc00000 19c00000 .....P..........
0070: 18900000 17400000 15f00000 14a00000 (dot)(dot)(dot)(dot)(dot)(at)(dot)(dot)(dot)(dot)(dot)(dot)(dot)(dot)(dot)(dot)
0080: 12c00000 3da00000 3ba00000 3a700000 ....=...;...:p..
0090: 39200000 37d00000 36800000 34a00000 9 ..7...6...4...
00a0: 00000000 ....

<Data> ------
Item 1 -- Length: 0 Offset: 7888 (0x1ed0) Flags: 0x00
Item 2 -- Length: 0 Offset: 7536 (0x1d70) Flags: 0x00
Item 3 -- Length: 0 Offset: 7064 (0x1b98) Flags: 0x00
Item 4 -- Length: 0 Offset: 6760 (0x1a68) Flags: 0x00
Item 5 -- Length: 0 Offset: 6408 (0x1908) Flags: 0x00
Item 6 -- Length: 0 Offset: 5936 (0x1730) Flags: 0x00
Item 7 -- Length: 0 Offset: 5632 (0x1600) Flags: 0x00
Item 8 -- Length: 0 Offset: 5280 (0x14a0) Flags: 0x00
Item 9 -- Length: 0 Offset: 4808 (0x12c8) Flags: 0x00
Item 10 -- Length: 0 Offset: 4504 (0x1198) Flags: 0x00
Item 11 -- Length: 0 Offset: 4152 (0x1038) Flags: 0x00
Item 12 -- Length: 0 Offset: 3680 (0x0e60) Flags: 0x00
Item 13 -- Length: 0 Offset: 3376 (0x0d30) Flags: 0x00
Item 14 -- Length: 0 Offset: 3024 (0x0bd0) Flags: 0x00
Item 15 -- Length: 0 Offset: 2552 (0x09f8) Flags: 0x00
Item 16 -- Length: 0 Offset: 2248 (0x08c8) Flags: 0x00
Item 17 -- Length: 0 Offset: 1896 (0x0768) Flags: 0x00
Item 18 -- Length: 0 Offset: 1424 (0x0590) Flags: 0x00
Item 19 -- Length: 0 Offset: 1120 (0x0460) Flags: 0x00
Item 20 -- Length: 0 Offset: 768 (0x0300) Flags: 0x00
Item 21 -- Length: 0 Offset: 296 (0x0128) Flags: 0x00
Item 22 -- Length: 0 Offset: 3552 (0x0de0) Flags: 0x00
Item 23 -- Length: 0 Offset: 3296 (0x0ce0) Flags: 0x00
Item 24 -- Length: 0 Offset: 3144 (0x0c48) Flags: 0x00
Item 25 -- Length: 0 Offset: 2976 (0x0ba0) Flags: 0x00
Item 26 -- Length: 0 Offset: 2808 (0x0af8) Flags: 0x00
Item 27 -- Length: 0 Offset: 2640 (0x0a50) Flags: 0x00
Item 28 -- Length: 0 Offset: 2400 (0x0960) Flags: 0x00
Item 29 -- Length: 0 Offset: 7888 (0x1ed0) Flags: 0x00
Item 30 -- Length: 0 Offset: 7632 (0x1dd0) Flags: 0x00
Item 31 -- Length: 0 Offset: 7480 (0x1d38) Flags: 0x00
Item 32 -- Length: 0 Offset: 7312 (0x1c90) Flags: 0x00
Item 33 -- Length: 0 Offset: 7144 (0x1be8) Flags: 0x00
Item 34 -- Length: 0 Offset: 6976 (0x1b40) Flags: 0x00
Item 35 -- Length: 0 Offset: 6736 (0x1a50) Flags: 0x00

*** End of Requested Range Encountered. Last Block Read: 110 ***

On Wed, Feb 07, 2007 at 03:00:20AM -0700, Ed L. wrote:
> How do I fix this 7.4.6 issue short of initdb?
>
> invalid page header in block 110 of relation "pg_statistic"

Take a copy of the file, then you should be able to truncate it.
There's also the zero_damaged_pages option, though I don't know whether
that applies here.

> I looked at the block via pg_filedump (included below), and it
> does not appear to me to be corrupted, so not sure what I would
> zero out, if anything.

I don't see anything odd expect that the header has both a lot of items
and a lot of free space...

> Items: 35 Free Space: 8032
> Length (including item array): 164

Have a nice day,
--
Martijn van Oosterhout <kleptog(at)svana(dot)org> http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.

"Ed L." <pgsql(at)bluepolka(dot)net> writes:
> How do I fix this 7.4.6 issue short of initdb?
> invalid page header in block 110 of relation "pg_statistic"
> I looked at the block via pg_filedump (included below), and it
> does not appear to me to be corrupted, so not sure what I would
> zero out, if anything.

Seems odd to me too ... but pg_statistic doesn't contain any
irreplaceable data. See if you can TRUNCATE it. If so, do
an ANALYZE to repopulate it. I don't recall if 7.4 has any
special hoops you'd have to jump through to truncate a system
catalog ...

regards, tom lane

On Wednesday February 7 2007 9:01 am, Tom Lane wrote:
> "Ed L." <pgsql(at)bluepolka(dot)net> writes:
> > How do I fix this 7.4.6 issue short of initdb?
> > invalid page header in block 110 of relation "pg_statistic"
> > I looked at the block via pg_filedump (included below), and
> > it does not appear to me to be corrupted, so not sure what I
> > would zero out, if anything.
>
> Seems odd to me too ... but pg_statistic doesn't contain any
> irreplaceable data. See if you can TRUNCATE it. If so, do
> an ANALYZE to repopulate it. I don't recall if 7.4 has any
> special hoops you'd have to jump through to truncate a system
> catalog ...

$ psql -c "truncate table pg_statistic"
ERROR: permission denied: "pg_statistic" is a system catalog

I suppose this means single-user mode is required? Any other
trick to do this during uptime?

TIA.
Ed

On Wednesday February 7 2007 9:01 am, Tom Lane wrote:
> "Ed L." <pgsql(at)bluepolka(dot)net> writes:
> > How do I fix this 7.4.6 issue short of initdb?
> > invalid page header in block 110 of relation "pg_statistic"
> > I looked at the block via pg_filedump (included below), and
> > it does not appear to me to be corrupted, so not sure what I
> > would zero out, if anything.
>
> Seems odd to me too ... but pg_statistic doesn't contain any
> irreplaceable data. See if you can TRUNCATE it. If so, do
> an ANALYZE to repopulate it. I don't recall if 7.4 has any
> special hoops you'd have to jump through to truncate a system
> catalog ...

Scheduled downtime finally arrived last night, and I tried
unsuccessfully to truncate this table:

postgres -D $PGDATA -O -o standalone_log mydb
> truncate pg_statistic;
> ^D

pg_ctl start
psql -d mydb -c "vacuum analyze"

The truncate showed no errors. The vacuum analyze showed the
same error in block 110 of the pg_statistic table. Any ideas
what I'm missing here? Is there another acceptable way to
truncate that table, perhaps "rm file; touch file"?

Ed

"Ed L." <pgsql(at)bluepolka(dot)net> writes:
> The truncate showed no errors. The vacuum analyze showed the
> same error in block 110 of the pg_statistic table.

Really!? Hm, I wonder if you have a reproducible problem. Would it be
possible for you to send me the physical pg_statistic file (off-list)?
I'm curious to see exactly what's wrong with it.

regards, tom lane