Chamikara Jayalath wrote:
On 1/9/06, Aleksander
Slominski <[EMAIL PROTECTED]
> wrote:
Chamikara
Jayalath wrote:
> Hi Alec,
>
> See my comments below.
>
> On 1/9/06, *Aleksander Slominski* <
[EMAIL PROTECTED]
> <mailto:
[EMAIL PROTECTED]>>
wrote:
>
> hi,
>
> my comments below.
>
> Chamikara Jayalath wrote:
>
>> ----- Original Message -----
>>
*From:* Chamikara Jayalath <mailto:[EMAIL PROTECTED]>
>> *To:*
[email protected]
>> <mailto:
[email protected]>
>> *Sent:* Friday, January 06, 2006 3:06 AM
>> *Subject:* [Sandesha2] Acknowledging policy
>>
>> Hi All,
>>
>>
It seems like we need to do some adjustments to our
>> acknowledging policy.
>>
>>
Currently acknowledging incoming application messages is
>>
done by the SandeshaInHandler. So acknowledging happens
>>
before the message is actually delivered to the service.
>>
>>>>But the message is received by the RMEndpoint and that
>> means we should acknowledge.
>>
>>
But it seems like we can provide a better quality
>>
reliability by not acknowledging till we actually invoke
>>
the service. This way we can guarantee the delivering of
>>
the message to the service even in the in-memory case.
>>
(I.e. if the client receive an ack he can be sure that
>> the service got actually invoked).
>>
>>>>Yes, but the problem is once the message is received by
>>
the RMEndpoint it is RMEndpoints responsibility to invoke
>>
the web service. So what we want >>is to improve the
>> reliability of the RMEndpoint.
>>>>IMHO we should not expect the initial sender to wait
>>
till the web service gets invoked for an acknowledgment.
>>>>Consider a scenario where we have 3 messages and the
>>
RMEndpoint manager in the destination receive 2 and 3 but
>>
not 1. We use INORDER >>invocation. Now we will not
>>
acknowledge for any of the messages since we did not
>>
receive message 1. This is not correct, because then the
>>
>>RMEndpoint manager in the client side will keep on
>> sending all the 3 messages again and again.
>>
>>
>>
>> yes, But if the server sends the messages and fail before
he
>> actually invoke the service, the client will proceed
believing
>> that the service got actually invoked. It is not important
>> weather the message got lost in the wire, or it got lost
within
>> the server, the result is the same (the service did not get
>> invoked). So the result is equal to acknowledging a
message the
>> server did not receive.
>> But performance wise what you say is very correct. If the
server
>> consume a long time to invoke the service, the client will
also
>> have to wait a long time for an acknowledgement.
>
> i am not sure but it looks to me that those are two different
> levels of reliability: reliable *message* delivery and reliable
> *service invocation*. the former is what i though WS-RM is for
and
> the latter is not supported in WS-RM unless some special QoS
> (Policy?) is used?
>
>
>
>
>
> What I was thinking abt was the usability. Most of the time the
> client may be unaware weather the service has implemented
persistent
> reliability or not. Also at times they may want to make sure that
the
> service actually got invoked before proceeding.Since we are able to
> provide that, why not add the feature.
AFAICT it will make Sandesha work in different ways from other WS-RM
implementations as it is an extension.
moreover in-most case this feature ("reliable invocation") is better
supported (and more interoperable) by having a service sending a
reliably response message (with "OK" or SOAP:Fault) ...
>
> i think that in case of in-memory storage not much can be done
for
> reliability but ask user to buy UPS :)
>
>
>
> Ya, that solves the prob. Bit costly though :-)
UPSes are cheap and without them server will not be very reliable anyway
if it keeps all data in memory - what if it needs to be rebooted ...
>
>
> IMHO the in-memory storage should be not a default in
Sandesha2 as
> it does not meet requirements of WS-RM instead some persistent
> storage should be a default one - maybe just using text files
to
> store message (as even embedded RDBMS may be overkill for
simple
> installations)?
>
>
>
> You have a point. But there may be many users who need an hi
> performing in-memory case. Those who need persistence (RDBMS or
file
> based) can do it by a single property change..
there is clearly a trade-off between reliability and performance -
however i am not sure if in-memory storage is good enough to call it
"reliable"? what should happen when there are many ack-ed but not yet
processed messages (for example with out-of-order messages) and server
goes down? those messages will be essentially lost when server (or
client) that is using in-memory storage crashed and is restarted (for
perfect reliability TX are needed to be able to rollback in-progress
unfinished invocations ...)
That is exactly the feature I'm talking abt. With this Sandesha2 will
not ack without invoking the message even in the in-memory case.
will it wait until invocation is finished? what if the service invoked
spins another thread to do actual computation/message processing
(typical for receiving one way messages when a service wants to return
fast to allow servlet container to close HTTP connection thread)? it
looks to me that this can only made reliable by combining WS-RM with
transactions (so no side effects and aborted invocations are rolled
back and be retried when server recovered by re-processing rolled back
message queues) ...
We can
also make this optional so that users who need acks quickly can get
them before invocation (this will be the default).
IMHO in-memory storage is not good choice for WS-RM impl and trying to
improve it may not be that useful ...
best,
alek
--
The best way to predict the future is to invent it - Alan Kay
|
