Bug #89992 rejoinInstance fails after a node goes MISSING but remove/add works
Submitted: 9 Mar 2018 15:48 Modified: 9 Mar 2018 15:52
Reporter: Kumar Verma (Platinum Quality Contributor) Email Updates:
Status: Verified Impact on me:
None 
Category:Shell AdminAPI InnoDB Cluster / ReplicaSet Severity:S3 (Non-critical)
Version:1.0.11, 5.7.20 OS:Any
Assigned to: CPU Architecture:Any
Tags: addinstance, missing, rejoinInstance, removeinstance, shell, timeout

[9 Mar 2018 15:48] Kumar Verma
Description:
After a node reaches a missing state due to unavailability of the instance (restart/network timeout/etc), rejoinInstance does not add it back. 

The mysqlprovision command contains the node itself as the group seed. 

How to repeat:
- Set up an InnoDB Cluster with 3 nodes with MySQL Shell 1.0.11 and MySQL Server 5.7.20
- Simulate an unavailable instance(close the VM and start back up)
- Ensure the state for the node is now MISSING and the cluster state is not NO_QUORUM (OK_NO_TOLERANCE or OK)
- Try to add the instance back using rejoinInstance. This should fail
- Remove instance and add back. This should succeed.

Notice the logs for the mysqlprovision command for rejoinInstance contains the node itself as the group seed.
[5 Apr 2018 10:57] Miguel Araujo
Posted by developer:
 
Hi Kumar,

I wasn't able to reproduce this issue using the latest Shell version: 8.0.11.

I'd like to ask you to double-check if you can reproduce the issue with the latest version and if so, to provide more information.