Re: 10.5 but not 10.4: backend startup during reindex system: could not read block 0 in file "base/16400/..": read only 0 of 8192 bytes

On Thu, Aug 30, 2018 at 01:25:00PM -0700, Andres Freund wrote:
> On August 30, 2018 1:24:12 PM PDT, Justin Pryzby <pryzby(at)telsasoft(dot)com> wrote:
> >On Thu, Aug 30, 2018 at 01:18:59PM -0700, Andres Freund wrote:
> >> Could you check if both of the proposed attempts at fixing the issue
> >> also solves your problem?
> >
> >Just checking that you're referring to:
> >
> >1) relcache-rebuild.diff
> >https://www.postgresql.org/message-id/20180829083730.n645apqhb2gyih3g%40alap3.anarazel.de
> >2) fix-missed-inval-msg-accepts-1.patch
> >https://www.postgresql.org/message-id/25024.1535579899%40sss.pgh.pa.us
>
> Right. Either should fix the issue.

Confirmed that relcache-rebuild.diff seems to fix/avoid the issue under 10.5,
and fix-missed-inval-msg-accepts-1.patch does so for HEAD (but doesn't apply
cleanly to 10.5 so not tested). That's for both the the new easy-to-hit
symptom and the pre-existing hard-to-hit issue affecting pg10.4 and earlier.

Thanks,
Justin