Pt-table-sync - second idle connection being timed out

Polyglot Projects Percona Toolkit
1

The company has been using pt-table-sync for a long time to sync tables from source → target - simple as can be no replicas involved no complications just:

pt-table sync [a few switches] sourceDSN destDSN

Source is a Galera cluster load balancer (Maxscale fronted). Recently I implemented an 900s/15m idle timeout on the source DSN’s load balancer and found that it broke the sync on large tables. I couldn’t see why until I spotted that pt-table-sync is making two connections to the DB - one that appears to be doing the sync queries and one that sits idle while the other one works. As some of the larger tables take more than 15m to complete, this seems to be causing pt-table-sync to throw errors.

I assume the second connection is some form of control connection but cannot see a way to either disable it or make it do any work between table syncs - i.e. a keepalive. Is this possible?

2

Why not just use basic, normal replication? This way your tables are instantly sync’d and you don’t need some external tool to manage.

3

I have looked at this a few times and have discounted it because there are reasons it doesn’t want to be in sync all the time. Yes there are ways to manage that but I don’t understand all the things that this interacts with and I’ll be honest, don’t want to.

This system has been in place for a very long time, and changing to binlog replication will almost certainly have unforeseen impacts and I’d much rather not have to mop up the mess if it does go horribly wrong somehow.