Horton

Re: prevent immature WAL streaming

Nathan Bossart <bossartn@amazon.com>

View thread View on postgr.es
From: "Bossart, Nathan" <bossartn@amazon.com>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Kyotaro Horiguchi <horikyota.ntt@gmail.com>, "andres@anarazel.de" <andres@anarazel.de>, "masao.fujii@oss.nttdata.com" <masao.fujii@oss.nttdata.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, "mengjuan.cmj@alibaba-inc.com" <mengjuan.cmj@alibaba-inc.com>, "Jakub.Wartak@tomtom.com" <Jakub.Wartak@tomtom.com>
Date: 2021-09-17T21:15:27Z
Lists: pgsql-hackers 
On 9/17/21, 1:32 PM, "Alvaro Herrera" <alvherre@alvh.no-ip.org> wrote:
> On 2021-Sep-17, Bossart, Nathan wrote:
>
>> I gave the patch a read-through.  I'm wondering if the
>> XLOG_OVERWRITE_CONTRECORD records are actually necessary.  IIUC we
>> will set XLP_FIRST_IS_ABORTED_PARTIAL on the next page, and
>> xlp_pageaddr on that page will already be validated in
>> XLogReaderValidatePageHeader().  Does adding this new record also help
>> ensure the page header is filled in and flushed to disk?
>
> That was the first implementation, a few versions of the patch ago.  An
> added benefit of a separate WAL record is that you can carry additional
> data for validation, such as -- as suggested by Andres -- the CRC of the
> partial data contained in the message that we're skipping.  I didn't
> implement that, but it should be trivial to add it.

I see.  IMO feels a bit counterintuitive to validate a partial record
that you are ignoring anyway, but I suppose it's still valuable to
know when the WAL is badly broken.  It's not expensive, and it doesn't
add a ton of complexity, either.

Nathan

Commits

  1. Fix determination of broken LSN in OVERWRITTEN_CONTRECORD

  2. Fix instability in 026_overwrite_contrecord.pl test.

  3. Fix typo

  4. Change recently added test code for stability

  5. Remove unstable, unnecessary test; fix typo

  6. Repair two portability oversights of new test

  7. Fix WAL replay in presence of an incomplete record

  8. Document XLOG_INCLUDE_XID a little better

  9. Revert "Avoid creating archive status ".ready" files too early"

  10. Avoid creating archive status ".ready" files too early

  11. Fix scenario where streaming standby gets stuck at a continuation record.

  12. Teach xlogreader to follow timeline switches

  13. Forbid using pg_xlogfile_name() and pg_xlogfile_name_offset() during