On Apr 25, 2007, at 9:42 AM, Simon Riggs wrote:
> On Thu, 2007-04-19 at 15:48 -0500, Thomas F. O'Connell wrote:
>
>> "If we take a backup of the standby server's files while it is
>> following logs shipped from the primary, we will be able to reload
>> that data and restart the standby's recovery process from the last
>> restart point. We no longer need to keep WAL files from before the
>> restart point. If we need to recover, it will be faster to recover
>> from the incrementally updated backup than from the original base
>> backup."
>>
>> I'm specifically confused about the meaning of the following phrases:
>>
>> "backup of the standby server's files" - Which files?
>
> the files that make up the database server:
> - data directory
> - all tablespace directories
>
>> "reload that data" - What does this mean in postgres terms?
>
> copy back from wherever you put them in the first place
>
> "that data" referring to the "files that make up the db server"
>
>> "last restart point" - What is this? Wouldn't it be able to restart
>> from the last recovered file, which would presumably occur later than
>> the last restart point?
>
> No, we don't restart file-by-file.
>
> http://developer.postgresql.org/pgdocs/postgres/continuous-
> archiving.html#BACKUP-PITR-RECOVERY
>
> "If recovery finds a corruption in the WAL..." onwards explains the
> restart mechanism. It's much like checkpointing, so we don't restart
> from the last log file we restart from a point possibly many log files
> in the past.
>
>> Does this mean make a filesystem backup of the standby server's data
>> directory while it's stopped, and then start it again with that data
>> and the restricted set of WAL files needed to continue recovery?
>
> No need to stop server. Where do you read you need to do that?
>
>> I'd like to see the language here converted to words that have more
>> meaning in the context of postgres. I'd be happy to attempt a
>> revision
>> of this section once I'm able to complete an incrementally updated
>> backup successfully.
>
> Feel free to provide updates that make it clearer.
>
>> Here's how I envision it playing out in practice:
>>
>> 1. stop standby postgres server
>> 2. [optional] preserve data directory, remove unnecessary WAL files
>> 3. restart standby server
>
> step 2 only.
>
> Clearly not an optional step, since its a 1 stage process. :-)
>
> --
> Simon Riggs
> EnterpriseDB http://www.enterprisedb.com
Well, this conversation made things a lot clearer, but I'm not sure
(yet) how to patch the docs. It seems like the original version is
written in general terms, whereas what our Q&A produces here is very
postgres-specific. I'll see if I can produce a version that would be
add clarity (for me).
--
Thomas F. O'Connell
optimizing modern web applications
: for search engines, for usability, and for performance :
http://o.ptimized.com/
615-260-0005
