Re: Completely broken replica after PANIC: WAL contains references to invalid pages

On 2013-11-05 12:03:19 -0800, Sergey Konoplev wrote:
> On Mon, Nov 4, 2013 at 1:48 AM, Andres Freund <andres@2ndquadrant.com> wrote:
> > On 2013-11-01 00:35:18 -0700, Sergey Konoplev wrote:
> >> On Tue, Apr 2, 2013 at 11:26 AM, Andres Freund <andres@2ndquadrant.com> wrote:
> >> > The attached patch fixes this although I don't like the way it knowledge of the
> >> > point up to which StartupSUBTRANS zeroes pages is handled.
> >>
> >> So, after half a year the same failure has happened again on the same
> >> replica, but now patched with the Andres' patch (9.2.4 + the patch)
> >> that was supposed to fix it.
> >
> > I think that's actually a different bug - what the patch was fixing was
> > the clog/subtrans errors, not the "reference to invalid pages".
>
> Can I do something to help fixing this particular bug? I backed up the
> relation and xlog files in case if someone need to dig it deeper.
>
> > I've submitted a fix for at least one cause causing that independently
> > from this bug which is included in 9.2.5
> > (17fa4c321ccf9693de406faffe6b235e949aa25f). Note thought that that fix
> > needed a fix iff using a cassert enabled build
> > (4da24f12e63313b7dbb6b3e3d0317e04045df636) which is *not* in 9.2.5.

As quoted above, I think the fix is in 9.2.5.

Greetings,

Andres Freund

--
 Andres Freund                       http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services