Re: page corruption on 8.3+ that makes it to standby

On Wed, Jul 28, 2010 at 7:02 AM, Simon Riggs <simon@2ndquadrant.com> wrote:
> On Tue, 2010-07-27 at 21:23 -0700, Jeff Davis wrote:
>
>> Both potential fixes attached and both appear to work.
>>
>> fix1 -- Only call PageSetLSN/TLI inside log_newpage() and
>> heap_xlog_newpage() if the page is not zeroed.
>>
>> fix2 -- Don't call log_newpage() at all if the page is not zeroed.
>>
>> Please review. I don't have a strong opinion about which one should be
>> applied.
>
> ISTM we should just fix an uninitialized page first, using code from
> VACUUM similar to
>
>  if (PageIsNew(page))
>  {
>    ereport(WARNING,
>        (errmsg("relation \"%s\" page %u is uninitialized --- fixing",
>                                                relname, blkno)));
>    PageInit(page, BufferGetPageSize(buf), 0);
>  }
>
> then continue as before.

As Jeff Davis pointed out upthread, you don't know that 0 is the
correct special size for the relation being copied.  In the VACUUM
path, that code is only applied to heaps, but that's not necessarily
the case here.

> We definitely shouldn't do anything that leaves standby different to
> primary.

Obviously.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise Postgres Company