On Sat, 2008-01-19 at 12:56 -0500, Tom Lane wrote:
> Erik Jones <erik@myemma.com> writes:
> > On Jan 19, 2008, at 2:26 AM, Simon Riggs wrote:
> >> Can you show us the messages, so we can understand the distribution of
> >> the pages?
>
> > All of the warnings are below. For tables that had multiple warnings
> > they seem to be for consecutive pages.
>
> Hmm --- that pattern sure looks like a smoking gun to me. Have you got
> any numbers about the likely relative sizes of the affected tables,
> and/or their growth rates while the rsync was in progress? What I am
> thinking is that the uninitialized pages probably represent the ranges
> of pages that were added to these tables while the base backup was going
> on. That could be partially confirmed, or conclusively disproven, if
> we had numbers about the table sizes and growth rates; but I think we
> could take it as a working hypothesis if we don't.
>
> If that is the case, then I can think of two explanations:
>
> 1. rsync misfeasance
>
> 2. the PITR replay started from some point later than that base backup
> started, and the missing pages are those that were added and filled in
> between.
>
> Theory #2 has two subtheories: 2A, you didn't follow the procedures
> properly, and 2B, you did but there's a bug in the procedures.
Since we already know (1) has occurred with the WAL file transfer, it
seems logical that the same cause could explain the uninitialized pages.
Two unrelated bugs are possible, but less likely.
But we need to check both...
--
Simon Riggs
2ndQuadrant http://www.2ndQuadrant.com