Today we experienced a similar outage. One of our JDBC connection pools
kept giving exceptions that it has reached the max pool size and the wait
timeout expired. However, all the connections in that pool were still
connected to the database and hadn't been used for 1.5 hours (according to
DBA). I tried flushing the pool using the Flush button in the web admin
console, but the connections were not closed and we kept getting the same
exception. Our DBA terminated all of the database connections from the DB
server, but the connection pool still gave exceptions about the pool being
full and max wait timeout. Then I changed the db username to something
invalid in the JDBC DataSource/Connection Pool, saved, ping, undo the change,
save, ping again. That seems to have reset the connection pool without
having to restart GlassFish because it is now working. We're not sure
what triggered this problem in the first place. We're running GlassFish
3.1-b43.
Ryan
[quote=vai_jpc]Howdy! Running 3.0.1 and every so often due to a network
break(?!) we can see a connection 'leak'. Its best observed by noticing the
server.resources.NameOfStoredProcPool.numconnused-current increment. If I
click 'flush' in the admin webpages or run asadmin flush-connection-pool
NameOfStoredProcPool both report success but neither change anything.
'numconnused-current' remains at the previous level.
Any suggestions?[/quote]
--
[Message sent by forum member 'rdelaplante']
View Post: http://forums.java.net/node/747522