MySQL Forums
Forum List  »  InnoDB clusters

InnoDB Cluster - avoid failover to node not in sync
Posted by: Sebastian Gentil
Date: September 14, 2018 04:56AM

Hello together,
I have been experimenting with Innodb Cluster for the last few days ... I am enthusiastic about the functionality!
But there is one thing that gives me a headache!

Test environment:
InnoDB cluster with 3 servers
single-master setup

Scenario:
- very high load on the master
- both slaves can not apply the relay logs fast enough and have a larger gap
- the master fails and an automatic failover to one of the slaves occurs

The application now accesses the new master, but it is not yet up to date because not all relay logs have been processed yet!
this means the application reads an outdated state of the data! This should not happen under any circumstances!

Is there a way to prevent the automatic failover to a slave that is not yet synchronous with the master?
No matter how long it takes, but the failover may only happen if the slave is synchronous!

thanks
Sebastian

Options: ReplyQuote


Subject
Views
Written By
Posted
InnoDB Cluster - avoid failover to node not in sync
56
September 14, 2018 04:56AM


Sorry, only registered users may post in this forum.

Content reproduced on this site is the property of the respective copyright holders. It is not reviewed in advance by Oracle and does not necessarily represent the opinion of Oracle or any other party.