Re: [bug fix] Cascading standby cannot catch up and get stuck emitting the same message repeatedly

From: pgsql-hackers-owner@postgresql.org
> [mailto:pgsql-hackers-owner@postgresql.org] On Behalf Of Amit Kapila
> I think it beginning of segment (aka the first page of the segment), even
> the comment indicates the same.
> 
> /*
> * Whenever switching to a new WAL segment, we read the first page of
> * the file and validate its header, even if that's not where the
> * target record is. ...
> ..
> */
> 
> However, on again looking at the code, it seems that part of code behaves
> similarly for both 9.2 and 9.3.

Yes, the code behaves similarly in 9.2 and later.  FYI, ValidXLogHeader() is called at two sites.  The earlier one checks the first page of a segment when the real target page is different, and the latter one checks any page including the first page of a segment.


> ..Because node3 found a WAL
> !  * record fragment at the end of segment 10, it expects to find the !
> * remaining fragment at the beginning of WAL segment 11 streamed from !
> * node2. But there was a fragment of a different WAL record, because !  *
> node2 overwrote a different WAL record at the end of segment 10 across !
> * to 11.
> 
> How does node3 ensure that the fragment of WAL in segment 11 is different?
> Isn't it possible that when node2 overwrites the last record in WAL segment
> 10, it writes a record of slightly different contents but which is of the
> same size as an original record in WAL segment 10?

That case is detected by checking the CRC value in the XLOG record header.

Regards
Takayuki Tsunakawa