Showing posts with label process. Show all posts
Showing posts with label process. Show all posts

Saturday, February 25, 2012

Message sent retry timeout

Hi

here is from BOL

>When a conversation is marked delayed, Service Broker performs the matching process again after a timeout period. Notice that failure to find a matching route is not considered an error.

for example it could happen when there is no route for a service.
the question is - what is this timeout. Is it configurable? on a system wide
basis ?

It will be great if this process is described in more details.

Best regards,

Leonid.The timeout period is not configurable. If you do want to force reclassification, however, you could trigger that by running ALTER ROUTE on some pre-existing route that basically does not change any attributes. The ALTER ROUTE would then have no effect, but it would trigger metadata change which causes Service Broker to retry classifying delayed conversations.|||

But what is the timeout is ? What is the overhead when SB trying to connect to a sevice with no route. As I understood SB will try to send a message every minute. Could you please clarify.

Leonid.

|||4 seconds that exponentially backs out to about a minute. But don't quote me on this.

Monday, February 20, 2012

Message Ordering

When I am receiving a batch of message from a queue to process, does it matter if I order by the queuing_order or the message_sequence_number? I want to make sure my messages process in the order in which they were put into the queue, which I know will happen if I process messages one-by-one, but want to make sure that the data I pull and process execute in the order in which they were received.

Tim

Either will do. RECEIVE projects the resultset in the order status, conversation_group, conversation_handle, queueing_order. I'd use the same order in my batch just for consistency (e.g. if the app logic is later moved to a CLR procedure processing a SqlDataReader, the order is unchanged).|||Sounds good, will do.

I did notice something the other day related to batch processing that I think could occur if you are not careful. It may be related to how I am handling the batch...let me know what you think:

I keep a convo open for ~ 90 seconds,or until a new message is sent. When I notice that a convo has gotten too old, I create a new one and process messages on that conversation. When I do this, I send an EndOfStream message that tells my procedures that it is time to end the conversation, and no new messages are sent on that older conversation. The potential problem I see with this is that when I create a new convo and send an EndOfStream message, the new convo starts processing messages right away, even maybe before the EndOfStream message has been processed, and all messages along with that convo processed. This, potentially, could mean that my messages are being processed out of order, which is really something that I can tolerate, but really don't want. Do you think it makes sense to try to add additional logic to prevent this from happening, or is there some other construct, such as a type of conversation lock I could use, to prevent this from happening?
Thanks,
Tim

Message OFF

In the query analyzer when I run the command
create database TestingDb
I can see the messages
The CREATE DATABASE process is allocating 0.63 MB on disk 'TestingDb'.
The CREATE DATABASE process is allocating 0.49 MB on disk 'TestingDb_log'.
I want to hide the message, how can I do that?
Thanks
We can not turn off those messages in query analyzer or management studio.
Thanks
Hari
"Rogers" <naissani@.hotmail.com> wrote in message
news:OyZC18InHHA.4424@.TK2MSFTNGP03.phx.gbl...
> In the query analyzer when I run the command
> create database TestingDb
> I can see the messages
> The CREATE DATABASE process is allocating 0.63 MB on disk 'TestingDb'.
> The CREATE DATABASE process is allocating 0.49 MB on disk 'TestingDb_log'.
> I want to hide the message, how can I do that?
> Thanks
>