Andrey Lizenko writes:
> Another point regarding collations: query with ORDER BY returns different
> record set on master and on replica if more than one column used:
When I saw Peter's reply, I figured he'd nailed it, and this seems to
be more evidence pointing in that direction. If the standby
Re-indexing didn't help. Will check via amcheck.
Another point regarding collations: query with ORDER BY returns different
record set on master and on replica if more than one column used:
Master:
> =# select date_started, symbol from raw.symbols_aggregates order by 1,2
> limit 5;
> date_starte
Andrey Lizenko writes:
> I'm observing strange behaviour on comparing ::text field with string while
> quering replica.
These symptoms seem consistent with the theory that the replica's index
for that column (symbols_aggregates_uniq) is corrupt. I think your
casts etc are just serving to defeat
On Wed, Jun 13, 2018 at 1:10 PM, Andrey Lizenko wrote:
> I'm observing strange behaviour on comparing ::text field with string while
> quering replica.
The first thing to do here is to install amcheck on the replica, and
verify that the indexes are consistent with the replica's own notion
of how
