BUG #14450: inconsistent results on async slave

From: rikard(at)ngs(dot)hr
To: pgsql-bugs(at)postgresql(dot)org
Subject: BUG #14450: inconsistent results on async slave
Date: 2016-12-06 08:57:36
Message-ID: 20161206085736.4360.39346@wrigleys.postgresql.org
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-bugs

The following bug has been logged on the website:

Bug reference: 14450
Logged by: Rikard Pavelic
Email address: rikard(at)ngs(dot)hr
PostgreSQL version: 9.3.14
Operating system: Linux
Description:

Before applying new version of the function to the database I wanted to
check if results of the function were the same

So we ran a query like this:

select a.id, original_fun(a.id), new_fun(a.id)
from table a
where coalesce(original_fun(a.id), MIN_DATE) != coalesce(new_fun(a.id),
MIN_DATE)

on all our async slave instances.
Strangely for small number of rows (6 out of 500k) this query returned
different results.
After running:

select original_fun(diff_id), new_fun(diff_id);

for one of those ids, the function returned the same result.

Is this expected behavior?

original_fun = plpgsql (volatile - should be stable)
new_fun = sql

What is even more strange, is that different results were like this:

id1, X1, null
id2, null, X2
id3, X3, null
id4, X4, null
id5, X5, null
id6, X6, null

data is only inserted and X_ represents min date of sum subset of data in
other tables.

Regards,
Rikard

Responses

Browse pgsql-bugs by date

  From Date Subject
Next Message tim.klicks 2016-12-06 15:07:49 BUG #14451: PostgreSQL server on iscsi disks
Previous Message Pavel Stehule 2016-12-05 18:57:43 Re: BUG #14446: make_date with negative year