MySQL Router Does Not Escalate Metadata Changes To Applications When Using Stop Group_replication
(Doc ID 2543233.1)
Last updated on FEBRUARY 27, 2020
Applies to:MySQL Router - Version 8.0 and later
Information in this document applies to any platform.
Should the Router notify the original connections that the node they are connected to has now become R/O, when they originally connected as R/W ?
We are currently testing how our application reacts to a change/loss of the Primary R/W node in the InnoDB cluster.
The application currently only connects to the R/W port of the Router. We can then look directly at the underlying mysql databases and see that the connections have been passed to the R/W node.
If we kill this Primary mysql node (kill mysqld/stop the host), we see the connections drop, a new node takes over as R/W, the Router reconfigures and new connections hit the new R/W node. All as expected.
If we do a 'stop group_replication' on the R/W mysql node instead , we see a new node take over the R/W role, and the Router reconfigures for the new setup. However, as mysqld has not stopped, the connections to the original R/W node do not get dropped, so do not reconnect to the new R/W node.
The Router HAS reconfigured (so new connections get routed correctly), but the original connections (which were originally routed to a R/W node) are now 'orphaned' (but still active) as this node is now R/O.
To view full details, sign in with your My Oracle Support account.
Don't have a My Oracle Support account? Click to get started!
In this Document