[ 
https://issues.apache.org/jira/browse/DBCP-427?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14259506#comment-14259506
 ] 

Phil Steitz edited comment on DBCP-427 at 12/29/14 2:54 AM:
------------------------------------------------------------

I like the idea here and the patch looks OK, except that I think this behavior 
should be configurable.  


was (Author: psteitz):
I like the idea here and the patch looks OK, except that I think this behavior 
should be configurable.  Also, a related feature that has been requested before 
and could be incorporated here is a method allowing client applications to 
explicitly mark connections as invalid.

> Examines 'SQLException's thrown by underlying connections or statements for 
> fatal (disconnection) errors
> --------------------------------------------------------------------------------------------------------
>
>                 Key: DBCP-427
>                 URL: https://issues.apache.org/jira/browse/DBCP-427
>             Project: Commons Dbcp
>          Issue Type: Improvement
>    Affects Versions: 2.0
>            Reporter: Vladimir Konkov
>            Priority: Minor
>             Fix For: 2.1
>
>         Attachments: DBCP-427_initial_impl.diff
>
>
> Affected cases:
> 1. Ability to detect severed connection without need for validity check (it 
> can be time demand).
> 2. Not all fatal errors detected with simple validation query or isValid() 
> call. For example: 
> Broken XA connections of MS SQL Server and Oracle Database (and may be other) 
> pass validation for both isValid() and simple validation query such as 'SELEC 
> 1 FROM DUAL' but throw fatal errors on data and/or transaction manipulation 
> calls. Such errors gone only after reconnect. In case of use DBCP for polling 
> application has no chance to cleanup broken connection.
> By providing infractructure for check thrown exceptions and mark 
> PoolableConnection as poisoned we can cleanup pool on validation phase.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to