On 10/22/2014 04:24 PM, Michael Paquier wrote:
> On Wed, Oct 22, 2014 at 1:53 PM, Heikki Linnakangas <hlinnakangas@vmware.com
>> wrote:
>
>> I think we should take a more wholesale approach to this. We should
>> enforce the rule that the server only ever archives WAL files belonging to
>> the same timeline that the server generates. IOW, the server only archives
>> the WAL that it has generated.
>
> Hm?! Would that be really back-patchable? There may be in the wild tools or
> users that rely on the fact a node archives segment files from all
> timelines.
Hmm, so it would be a tool or user that manually copies a file to the
pg_xlog directory of a standby server, and expects the standby to
archive the file after promotion. That seems a bit far-fetched, although
I've seen people do strange things. I think it would be acceptable as
long as we document the change in behavior in the release notes. I don't
have much hope that we'll ever be able to nail down the correct behavior
with the current approach.
- Heikki