Re: 9.3 Beta1 status report - Mailing list pgsql-hackers

From Heikki Linnakangas
Subject Re: 9.3 Beta1 status report
Date
Msg-id 51766F94.80706@vmware.com
Whole thread Raw
In response to Re: 9.3 Beta1 status report  (Bruce Momjian <bruce@momjian.us>)
Responses Re: 9.3 Beta1 status report
List pgsql-hackers
On 22.04.2013 23:06, Bruce Momjian wrote:
> On Mon, Apr 22, 2013 at 10:11:48PM +0300, Heikki Linnakangas wrote:
>>> E.1.3.2.1. Write-Ahead Log (WAL)
>>>
>>>     Store WAL in a continuous stream, rather than skipping the last 16MB segment every 4GB (Heikki Linnakangas)
BACKWARDCOMPATIBLE BREAK
 
>>>
>>>     Restructure WAL files to better handle timeline changes during recovery (Heikki Linnakangas)
>>>
>>>     Restructure WAL files to use a more compact storage format (Heikki Linnakangas)
>>
>> Can you clarify which commits these came from? The first one is
>> clear (dfda6eba), and I think the 3rd covers commits 20ba5ca6 and
>> 061e7efb1. But what is that second entry?
>
> Frankly, I found the WAL and timeline commits all over the place and
> could hardly make sense of it.  I tried to collapse entries into
> meaningful items, but I need help.  Can you suggest changes?

Ok:

* Don't skip the last 16 MB WAL segment every 4GB, with filename ending 
in FF (Heikki Linnakangas) BACKWARD COMPATIBLE BREAK

* Change WAL record format, allowing the record header to be split 
across pages. The new format is slightly more compact. (Heikki Linnakangas)

In "Source Code" section:

* Use a 64-bit integer to represent WAL positions (XLogRecPtr), instead 
of two 32-bit integers. (Heikki Linnakangas)


Do we usually repeat the changes listed in the backwards compatibility 
section later, in the "Changes" section? If not, then instead of the 
first two items above, let's just have these in the 
backwards-compatibility section:

* The WAL file numbering has changed to not skip WAL files ending with FF.
  If you have e.g backup / restore scripts that took the skipping into  account, they need to be adjusted.

* The WAL format has changed.
  Any external tools that read the WAL files need to be adjusted to  understand the new format. The new xlogreader
facilityhelps  writing such tools.
 

- Heikki



pgsql-hackers by date:

Previous
From: Pavan Deolasee
Date:
Subject: Re: Couple of issues with pg_xlogdump
Next
From: Andres Freund
Date:
Subject: Re: Couple of issues with pg_xlogdump