Le 22/12/2012 00:32, Marina Vatkina a écrit :
> On 12/21/12 3:03 PM, Laird Nelson wrote:
>>
>> Since flushing the connection pool is a known bug (i.e. there's no
>> way from the admin console to flush a pool without causing an error;
>> see http://java.net/jira/browse/GLASSFISH-14130 for an example; this
>> is true even of our initialized pools on 3.1.2.2), can I somehow tell
>> GlassFish to dump and refill the pool at the first sign of connection
>> trouble? I didn't see anything obvious in
>> http://docs.oracle.com/cd/E26576_01/doc.312/e24928/jdbc.htm#gharo.
>
> Hmmmm ... connection validation should be documented...
I would also appreciate such a documentation : have the same exact problem
where a restart of the database server (this is very unlikely but hey :
it may
happen in HA environments in case of master/slave switch for example)
generates domain connections errors looong after the DB server is up.
I presume these are caused by staled pooled connections.
I tried to tweak some validations parameters (validation by a SQL query
for example), but could not get to something to work.
Any advice ?
Regards,
M. Maison