On 2016-02-08 16:32:22 +0000, bear2k(at)mail(dot)ru wrote:
> We are seeing a very strange situation - table pg_constrint has a duplicate
> row for particular primary key constraint which is not accessible via
> indexes but is visible during segment scan. There wasn't any specific
> manipulation with catalog data, however the database is used with standby
> node, so switchover from one instance to another is performed from time to
> time.
Did your have any failovers or restore from backup, at some point in the
history of the cluster? If so, what were the exact procedures you used?
Greetings,
Andres Freund