Bug #93466 MySQL router
Submitted: 4 Dec 2018 10:32 Modified: 10 Aug 14:48
Reporter: Shams ur Rahman Saeed Email Updates:
Status: Can't repeat Impact on me:
Category:MySQL Router Severity:S4 (Feature request)
Version:0.13 OS:Any
Assigned to: MySQL Verification Team CPU Architecture:Any

[4 Dec 2018 10:32] Shams ur Rahman Saeed
Dear Team,

I have tested MySQL router: it gives error while forwarding read queries to MySQL InnoDB cluster with only one database server left alive (only one server is up it automatically owns read write role) while other members of the cluster are failed.
In such cases queries with port 6447 (read only port) causes error to the application and can not proceed request. 

I have gone through lots of internet search but was unable to find a solid solution for such scenario. 



How to repeat:
*: MySQL InnoDB cluster is running with 3 nodes.
*: MySQL router router is deployed with ports 6446 read/write and 6447 read only.
*: Two of MySQL servers are gone and only one is left.
*: The only left server has the role of read/write and accepts queries with port 6446 only.

*: MySQL router cannot process queries with port 6447 (read only) and causes service interruption.

Suggested fix:
 I would like to suggest if any feature is to be added into MySQL router to perform read queries even with only one available server in cluster.

The router should check the performance_schema.group_repilcation_members of the available serve, in case it cannot get connect with any SECONDARY (read only) server.
[10 Aug 14:48] MySQL Verification Team

I cannot reproduce this behavior with 8.0.34/8.1.0.

please retest with latest verosions of server and router