oracle rdbms connector

[Cantor, Scott]

> To summarize in case I did not explain it well enough.  The IDP is working and
> answering requests.  A DB connector goes offline and the IDP stops
> answering requests while it waits for the DB connector to come back.  Once
> the DB connector comes back the IDP again starts answering requests.

Well, eventually all the threads will hit the connector logic and wait for the database, that's why it's hung. You have to guarantee the behavior of the connectors you configure. Any timeouts involved have to be set on that driver, they don't come from Shibboleth.

The IdP cannot be hung otherwise unless your container is badly broken. If there were request threads available, it should still respond to requests that don't hit the connector like the old status handler that just returns "ok".

> My comment about maxActive="0" was in reference to setting the pool to 0
> and only initiating a DB connect when one was needed.  The tomcat/dbcp
> documentation say that "0" means no limit and not none.

Well, this problem lies with that layer, not here. I would suggest you find out what the error message from Oracle means and maybe start with that. Then you might ask somebody that supports DBCP or Tomcat to address what would be used to get around whatever causes that error condition. 

-- Scott