Be like Project Farm, and test that. 😀 For example, insert records with all sorts of non-ASCII characters (i.e. letters with diacritic marks, œ, æ, other currency symbols etc) into "SELECT text_fld FROM very_big_table_ ORDER BY text_fld" into a csv file on each server, and then take the md5sum.
Testing various regex predicates in WHERE clauses is also a good ise.
If they're not the same, well, you have your answer: it's a problem. If they _are_ the same, then *either* it's not a problem *or* your test data is insufficient.
Hi @Ron,
What if collation is glibc independent i.e inbuilt C.UTF-8? still do we have any chance of inconsistency?
Regards,
OM
Hi Team,
Greetings!!
I have 3 node patroni postgresql cluster in which primary and replica is running on Ubuntu 20 and third one is running on Ubuntu 24 that is in DR region.
So if we do the switch over on DR and then switch back to primary region, is there any chance of data corruption due to OS version mismatch.
The problem is that corruption in text indices can happen due to collation changes from gnu libc version upgrades..
-- Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
