From: | Alex Derbes <acd(at)brillig(dot)org> |
---|---|
To: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | pgsql-bugs(at)postgresql(dot)org |
Subject: | Re: Corrpupt Data File |
Date: | 2004-02-26 17:22:01 |
Message-ID: | Pine.LNX.4.44.0402260909040.31950-100000@m1.brillig.org |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
for the annals of postgres maintenance I got my DB running by using DD to
copy the blocks before the corrupt block, then add block from /dev/zero,
then add the remaining blocks from the file. I discovered the block size
by looking at he output of pg_filedump. To find the offending file select
rel_filename from pg_class where relname = TABLE_NAME.
My original error was : "ERROR: Invalid page header in block 57879 of
orders" -- thus the bad block is #57879
cp 309928249 orders.bak
dd bs=8192 count=57879 if=309928249 of=out
dd bs=8192 count=1 seek=57879 if=/dev/zero of=out
dd bs=8192 skip=57880 seek=57880 if=309928249 of=out
mv out 309928249
Many thanks to Tom for his quick reply & Good luck to all others.
Thanks,
-Alex
On Thu, 26 Feb 2004, Tom Lane wrote:
> Alex Derbes <acd(at)brillig(dot)org> writes:
> > deleting the indexes from the table did not help. Is there a tool or
> > method that that I could use to simply skip the bad data and at least
> > recover the good blocks?
>
> Just zero out the broken page, eg with dd from /dev/zero. Note it
> might be a good idea to shut down the postmaster while you do so,
> else there could be a copy of that page in Postgres shared buffers,
> which'd mask your attempted fix.
>
> regards, tom lane
>
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
acd(at)brillig(dot)org - acd(at)cwru(dot)edu
f720.920.1148 - c415.572.1669
http://www.brillig.org/~acd
01 1010 100
From | Date | Subject | |
---|---|---|---|
Next Message | Josh Berkus | 2004-02-26 17:40:03 | Re: Known issue with Reindex-based corruption? |
Previous Message | Tom Lane | 2004-02-26 14:00:17 | Re: Corrpupt Data File |