Limitations for the Auto failover feature in ScaleArc for Oracle 3.10
1. Connections opened before Auto-failover flag is enabled are not switched over after failover. These connections are still served from original R+W server which is now demoted to standby after failover (IDB-10131).
2. During Auto failover, if a client connection is in transaction then ScaleArc will not switch the connection until the transaction is committed or roll backed.
3. During Auto failover, if ScaleArc is in between reading a response from the database server then ScaleArc will not switch the connection until the entire response is read. However, if the server goes down, before the entire response is completely read from the server, then Scalearc closes such connections.
4. If application is connection pooled, ScaleArc will not be able to detect when a connection is getting reused by the application. Therefore, whenever failover happens, ScaleArc will replay all the alter session commands that it has stored for that connection.
5. We will treat queries with pending fetches as “in progress queries” and after failover consecutive queries can fail with new database server connections. Application will get the same error message that has been generated by the database server after failover.
6. Currently only one Read/Write server per cluster is supported in ScaleArc.
7. ZDM will not work for those users which are not configured with ScaleArc.
8. Currently only a single SID or ServiceName can be specified per DB server when creating new cluster.
9. All DB servers which belong to the same cluster must have the same root user configured.
10. Existing connections on primary server do not switch over to newly promoted primary server even after changing the role of the server. (IDB-10669).
11. After failover, R/W split does not work for the existing connections, i.e., read query does not go to the recently demoted primary server for existing connection. R/W split functions as expected for all the new connections generated after failover.