Bug #96606 InnoDB Cluster - Prevents reads on Failed Primary Node
Submitted: 21 Aug 2019 16:58 Modified: 30 Oct 2019 12:36
Reporter: Pedro Pinheiro Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Server: Group Replication Severity:S4 (Feature request)
Version:8.0 OS:Any
Assigned to: CPU Architecture:Any
Triage: Needs Triage: D5 (Feature request)

[21 Aug 2019 16:58] Pedro Pinheiro
Description:
Use Case:
3 Node Single-Primary Cluster

Event:
Current Primary changes (Failed, leave the cluster, intentional  changed to another node)

OnGoing:
Application is connected to the Primary node, and as no way to know the primary node did change (it's only performing reads).

Problem:
How to prevent the application from continue to read from the failed primary? We don't want to read from a node that is not part of the cluster.
 Is there any setting on Group Replication or MySQL Router, to kill all the ongoing connections to the failed node? Assuming it's working and but not in the cluster. Or any other way to prevent the stale readings?

Note:
This would be a very desired feature to make InnoDB Cluster a reliable High Availability Solution.

How to repeat:
please check the description.
[22 Aug 2019 6:40] Umesh Shastry
Hello Pedro,

Thank you for the feature request.

regards,
Umesh
[30 Oct 2019 12:36] Nuno Carvalho
Posted by developer:
 
Implemented on WL#12895: "Group Replication: Add OFFLINE_MODE to group_replication_exit_state_action" released on 8.0.18