| Ken Hygh 2007-03-16, 1:28 am |
| khash23@yahoo.com wrote:
> I just want to get some general understanding for the expected behavior of WebSphere when it comes to database connection pooling.
>
> Consider a scenario where an application connecting to a database reaches its max # of connection that's defined in WAS. What should be the behavior of the application server?s connection pooling and the applications using these pool to connect to the
database. We have situation where the App Server crashes under this situation.
>
> Also, if the database has locking issues, what happens to the connections. Do they get released to the pool after certain time, or stay connected in which case you run out of connection objects.
>
> Any help would be appreciated.
>
I'll give my observations, perhaps prompting someone who actually knows
the code to respond :-)
Note that connection pooling has changed multiple times over WAS's
lifetime, so what I used to observe may no longer be the case :-0
Assuming that you have not configured the connection pool to expand
beyond its max....
When the pool is fully occupied, and your application does a
getConnection(), that thread will block until a connection is freed up,
or a timeout is reached. Now, depending on what the connection is
actually doing when some delay (the DB locks up, goes away, takes a
loooong time, etc) either an immediate exception will be thrown, or
another timeout will occur. Much of the variability here depends on the
database's jdbc driver and is outside of WAS's direct control. If the
underlying physical connection fails, the pool will attempt to destroy
it and recreate it.
The only time I've seen WAS crash in a database call is when the jdbc
driver uses native code (.dll or .so) which causes a crash. I haven't
seen it happen with a pure Java driver.
Ken
|