On Sun, 23 Jul 2000 08:13:24 -0400, dan benanav <[EMAIL PROTECTED]>
wrote:
>This is a good point Rickard. However in a single server system I would
>probably retry less often than in a clustered system. Wouldn't it be better to
>let the container handle this? Also is it always the case that one should
>retry on a RemoteException? RemoteExceptions can be generated in cases other
>than server failure can't they?
There's a subtle issue here: a RemoteException might not be a
RemoteException even though it looks like one. Why? Because you can get
subclasses of RemoteException! I'm not sure about RMI/IIOP, but if you
get ConnectException or ConnectIOException with RMI/JRMP it is pretty
safe to retry.
However, this *is* something that the container should know best, and be
able to handle transparently. If this happens in a two+ step process,
i.e. bean calling bean in another server, then it might be more
problematic to retry.
Yes, it is a tricky one.
>> >4) In single server mode you can assume pessimistic concurrency is
>> >used. In a clustered mode you must assume optimistic concurrency is
>> >used.
>>
>> Yes, since there is no global locking scheme in place. This is an
>> important difference that may have serious implications on the
>> applications semantics.
>
>And unfortunately Weblogic offers no advice on this. I am sure many
>programmers have relied on pessimistic concurrency for correctness of their
>programs.
Absolutely.
Question: is it bad that there is no advice, or that programmers rely on
this feature to be available?
Followup: by relying on this, aren't the components less portable?
And: is portability overall still a very low-priority issue??
>Doesn't impact the correctness but it might impact performance. So you may be
>relying of dbIsShared for performance. When you switched to clustering if you
>relied on dbIsShared=false your might take a big performance hit. Clustering
>is supposed to enhance salability not decrease it. Perhaps this is another
>warning weblogic should give. Don't rely on dbIsShared for performance if you
>plan to switch to a clustered system.
Correct, although I think there are caching options even with Weblogic
which allows this (Weblogic+Javlin for example?).
Nevertheless, things aren't as easy as they should be, but neither are
they as hard as they could be. Ah well.
/Rickard
--
Rickard �berg
Email: [EMAIL PROTECTED]
http://www.telkel.com
http://www.jboss.org
http://www.dreambean.com
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
