From: pgsql-admin-owner@postgresql.org [mailto:pgsql-admin-owner@postgresql.org] On Behalf Of hydra Sent: Friday, June 05, 2015 12:33 AM To: pgsql-admin Subject: Re: [ADMIN] replication consistency checking
On Fri, Jun 5, 2015 at 4:40 AM, Sergey Konoplev <gray.ru@gmail.com> wrote:
The primary symptom of this corruption is rows that:
are present on the master, but missing on the replica
have been deleted on the master still appear to be visible on the replica
have been updated, and their old versions appear alongside the new, updated versions on the replica
How can I verify whether I already have this corruption? There is no known way to identify that the issue has affected a standby in the past but comparing the data from the primary with the standby.
So hands up who still think PostgreSQL doesn't need some way of checking the data consistency between master-standby? :)
My hand is up.
From the wiki page that you referenced you forgot to quote this:
“This is an issue, discovered Nov. 18, 2013., which can cause data corruption on a Hot-Standby replica when it is (re-)started, by marking committed transactions as uncommitted. This issue is fixed in the December 5th 2013 update releases.”
So, what’s your point? Yes, any software can (and does) have bugs. What’s important is how quickly the bug is discovered and fixed.
You wish to write a utility that compares data on 2 different clusters – sure, by all means, but I believe time could be better spent on something else.
But, how is to say that this utility will be 100% error-free?
