On Wed, 2021-02-24 at 12:50 +0100, Václav Steiner wrote: > We have three servers running postgres 9.6, master and two slaves feeded by > streaming replication. > On of those slaves we are getting different query results. The replica was > recreated from scratch, but problem persists. > > If set enable_indexscan='off' and set enable_bitmapscan='off' we get correct > result. > It looks like problem with some index...? > > Can anybody saw such behaviour?
If the problem persist when you rebuild the standby, I suspect that the servers are running different C libraries or C library versions, so that the collations with the same name behave slightly differently. That could lead to the observed behavior with indexes on string data types. Make sure to use the same C library version on both systems. Yours, Laurenz Albe -- Cybertec | https://www.cybertec-postgresql.com
