[ https://issues.redhat.com/browse/TEIID-5925?page=com.atlassian.jira.plugi... ] Steven Hawkins commented on TEIID-5925: --------------------------------------- Admittedly there's not a lot of detail in the stackoverflow post, but as described there the pg source is acting as a materialization of the excel data. Only if pg is not available, then consult the excel source. So cache first and if not available then source - which looks like a different on error action. That's a more general failover scenario. It's made more complicated if you are dealing with heterogeneous sources (pg/excel) as the source metadata won't necessarily line-up across failover targets (multi-source has the same issue) . Ideally we lean on source support where possible to handle failover. There were even a few JDBC projects over the years to provide generic clustering/failover at the driver level. I'd go so far as to say in most homogeneous scenarios involving sql/nosql sources there would be source support of some kind. I see the appeal of fleshing out a general failover feature as that leads to other source query modes, such as parallel execution to obtain the first/fastest result, etc. However if it's primarily for what looks to be a materialization scenario, I'd start there. -- This message was sent by Atlassian Jira (v7.13.8#713008)